# AI-Powered Hacking: How to Protect Yourself

URL: https://www.mattsenter.com/blog/ai-is-making-hacking-easier-how-to-protect-yourself

*AI-powered hacking makes established security habits more urgent because attackers can automate convincing scams and exploit unpatched software faster. Start with a reputable password manager that generates a unique password for every account, and keep its recovery information securely stored offline. Protect your primary email account with a passkey or strong multi-factor authentication, since your inbox often controls password resets elsewhere. Enable automatic updates, review application permissions, and independently verify unexpected requests using contact information you already trust. For financial accounts, enable transaction alerts, freeze your credit with all three major bureaus, and report unauthorized activity promptly. Keep encrypted copies of important records and a separate backup you can reach if your vault becomes unavailable. AI can help research purchases, but granting it payment credentials and unrestricted spending authority creates additional risks. These measures reduce exposure; they cannot guarantee protection, so prepare a recovery plan before an incident.*

*Disclaimer: This article offers general information, not professional cybersecurity, financial, or legal advice. You are responsible for your own security decisions, including actions you take or choose not to take. I accept no responsibility or liability for any resulting losses or damages.*

Last night, my mother-in-law texted me after reading a [new Google report about AI-powered cyberattacks](https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold). She had a pretty reasonable question:

> "How do we ordinary peeps protect our sensitive personal info?"

My initial response was that the advice hasn't changed all that much. Don't install random applications, be suspicious of links in unexpected emails, use strong passwords, and enable two-factor authentication. What has changed is the scale and speed at which attackers can operate.

Google's November 16, 2026 report describes hackers using autonomous AI systems to discover vulnerabilities, develop exploits, and conduct multi-stage attacks with significantly less human intervention. Some of these systems can examine software updates, identify the vulnerabilities being patched, and develop working attacks against systems that haven't been updated yet. Other tools help criminals automate phishing campaigns, create convincing fake identities, and scale their operations far beyond what would previously have been practical.

Much of the report concerns attacks against businesses and their infrastructure rather than individual consumers. That's an important distinction, but it isn't particularly comforting when those businesses are holding our personal information.

The good news is that protecting yourself doesn't require understanding how these attacks work. Most of the recommendations are things we've been telling people to do for years. The challenge has always been getting people to actually follow them.

My exchange with my mother-in-law turned into a longer conversation about password managers, banking information, and the practical things an ordinary person can do to stay safe. I figured it was worth writing down.

## Stop trying to remember all your passwords

The first thing I recommended was getting a password manager, specifically [1Password](https://1password.com/).

My mother-in-law responded:

> "Thanks...never heard of 1Password. Will look it up. Long password? Good strategy. But how do I remember it with my aging brain?"

That's actually the entire point of a password manager. **You're not supposed to remember your passwords anymore.**

A password manager is a secure, encrypted vault that remembers your usernames and passwords for you. When you create an account on a website, it can generate a long, completely random password. When you return to that website, it automatically fills in your login information.

You don't need to know what that password is. It could be 30 random characters long, and you would never need to memorize or type it. In fact, I couldn't tell you what the passwords are for almost all of my own accounts.

Instead, you remember one strong password that protects your password manager. Ideally, that's a long, unique passphrase made from several randomly chosen, unrelated words, rather than something predictable like your dog's name and birthday. On supported devices, you can usually unlock the manager with Face ID or your fingerprint, although you'll occasionally need your account password.

1Password has another security feature called a Secret Key, which I'll explain shortly. You don't have to memorize that either.

The important part is that every account gets a different password. If some random shopping website gets hacked and exposes its customer credentials, the password you used there is worthless to someone trying to log into your bank or email.

You can also use a password manager on multiple devices. Your phone, laptop, and desktop can all have access to the same encrypted vault. You don't have to remember passwords when switching between them.

1Password isn't the only good option. [Apple Passwords](https://support.apple.com/guide/passwords/welcome/mac) is built into Apple's ecosystem and is free. [Bitwarden](https://bitwarden.com/) is another well-established option with a free tier. I particularly like 1Password, but the important decision is to start using a reputable password manager rather than continuing to invent and reuse your own passwords.

For someone new to this, I'd suggest starting with your email and financial accounts, then gradually updating other passwords. There's no reason to spend an entire afternoon changing hundreds of credentials.

### Yes, you can write down your master password

For years, we've been told never to write down passwords. That's generally sensible advice when we're talking about leaving passwords on sticky notes attached to monitors, but it ignores an important distinction between physical and digital security.

The United Kingdom's [National Cyber Security Centre also says writing passwords down can be done safely](https://www.ncsc.gov.uk/collection/top-tips-for-staying-secure-online/password-managers). Its guidance specifically discusses journals and password books as an arrangement that can suit people who receive regular help from family members managing their online accounts. It emphasizes choosing strong passwords, considering who can physically access the record, and checking whether a service's terms permit writing passwords down. Simply hiding the book may not be enough, and paper still requires you to recognize fraudulent websites yourself.

A criminal sitting halfway around the world can attempt to steal your online credentials using phishing, malware, compromised websites, or automated attacks. That same criminal cannot simply reach into a locked safe in your house and retrieve a piece of paper.

Physical security and digital security involve different threats. A piece of paper stored in a secure location is not exposed to the same remote attacks as an online account or an unencrypted file sitting on your desktop.

That doesn't make paper inherently safe. Someone with physical access could steal it. It could be lost, destroyed in a fire, or photographed by someone who shouldn't see it. The point is that those risks are different and can be managed with appropriate physical safeguards.

This is also [1Password's own recommendation](https://support.1password.com/emergency-kit/): print your Emergency Kit, write your account password on at least one printed copy, and store it safely. The company specifically suggests a safe deposit box as one option. I recommend a similarly secure location, such as a locked home safe. The kit includes your Secret Key, which is an additional randomly generated secret used to protect your encrypted vault. Together, your account password and Secret Key are extremely sensitive, so don't leave that document in an unlocked drawer or email yourself a photograph of it.

If you use two-factor authentication on your 1Password account, make sure you've also planned how to recover that access if your phone is lost. Follow the company's recovery instructions rather than assuming the Emergency Kit alone covers every situation.

A securely stored recovery document isn't an embarrassing workaround for somebody who can't remember passwords. It's a deliberate backup mechanism. You may never need it, but if you lose your phone or forget your account password, you'll be grateful you prepared it.

For couples and families, [1Password Families](https://support.1password.com/family-recovery-plan/) also offers shared vaults and account-recovery capabilities. You can keep your personal passwords private while sharing things like household utility accounts. With the right setup, a designated family organizer can help another family member recover access.

The goal is to make your digital life safer and easier, not give you another complicated system to worry about.

## Isn't a password manager a single point of failure?

A couple of months ago, I [asked Reddit why so many people resist using password managers](https://www.reddit.com/r/askanything/comments/1wuc63g/why_are_some_people_so_averse_to_or_otherwise/). I received a lot of responses, and one objection came up repeatedly: putting all your passwords in one place creates a single point of failure.

It's a legitimate concern. If someone compromises your unlocked password vault, they could potentially access every account stored inside it. There's no reason to pretend otherwise. The question is whether that risk is greater than the risks introduced by whatever alternative you've devised for yourself.

One commenter told me they refused to use a password manager because they kept their passwords in an encrypted file. Another kept a physical notebook locked in a wall safe. Others described personal formulas for generating passwords based on the websites they visited.

I found the encrypted-file response particularly interesting. That person had effectively created their own password manager, only without the features and protections offered by dedicated software.

The physical notebook deserves a little more consideration. Unlike a file stored on an internet-connected computer, a notebook sitting in a locked safe can't be directly accessed through a remote software vulnerability. For someone with very few accounts, it may be a perfectly reasonable system.

The problem isn't that paper is inherently insecure. It's that paper becomes increasingly impractical as the number of accounts grows.

If you have a hundred online accounts, are you going to write down a unique, randomly generated password for each one? What happens when you're traveling and need to sign in somewhere? Do you carry the notebook, potentially losing it or exposing it to theft? What happens when you need to update a password on your phone and the notebook is at home?

The notebook also doesn't provide any protection against typing a password into a fraudulent website. A password manager can recognize which website a credential belongs to and generally won't offer to fill it into an unrelated domain.

If a notebook works for your particular circumstances, I'm not going to tell you it's inherently unsafe. I am going to suggest that a properly configured password manager solves considerably more problems for anyone who regularly uses modern online services.

Personal password formulas are especially problematic. Taking a familiar password and modifying it based on the website might produce different passwords, but different doesn't necessarily mean unpredictable. Once an attacker discovers the pattern, the security of every password generated using it may be compromised.

Professional password managers approach the problem differently. They generate independent, random credentials and protect them inside an encrypted vault.

With 1Password, your vault is encrypted on your device before being uploaded to its servers. The company doesn't possess the information necessary to decrypt it. Your account password and a separately generated Secret Key are combined to protect your vault. The Secret Key is generated locally and isn't available to 1Password, meaning someone who steals encrypted vault data from the company's servers doesn't automatically have what they need to decrypt it.

This is a sophisticated security architecture, which you can read about in [1Password's security design white paper \[PDF\]](https://1password.com/files/1Password-White-Paper.pdf).

Of course, password managers aren't invulnerable. In 2022, [LastPass suffered a major security breach](https://blog.lastpass.com/posts/notice-of-security-incident) in which attackers stole backups containing customers' encrypted vaults along with some unencrypted information, including website addresses. That created opportunities for attackers to attempt offline password cracking, particularly against vaults protected by weak master passwords.

That's a serious incident, and I wouldn't dismiss anyone's concern about it. The lesson, however, isn't that we should abandon password managers. It's that the security architecture of the manager matters, along with how you configure and use it.

There's also a difference between someone stealing an encrypted copy of your vault and someone compromising your computer while that vault is unlocked. Malware running on your computer may be able to steal credentials after they've been decrypted for use. A password manager isn't a substitute for securing your devices.

Losing access to your vault is another legitimate concern. That's why I recommend having the manager installed on more than one trusted device, keeping recovery information safely stored offline, and configuring recovery options before you need them.

The [National Institute of Standards and Technology (NIST)](https://pages.nist.gov/800-63-4/sp800-63b.html) explicitly supports the use of password managers, noting that they encourage stronger, unique passwords and reduce the burden of remembering credentials.

I think the supposed single point of failure is often misunderstood. A password manager does concentrate some risk, but it also dramatically reduces the number of places where your own password habits can create vulnerabilities.

Instead of reusing the same password across 50 websites, you're using 50 independent passwords. A breach at one website no longer compromises your credentials everywhere else.

I'm a software engineer, and I'm not going to pretend that a homegrown system I invented is automatically safer than one developed specifically to protect credentials. I could build a password manager, but then I'd also be responsible for maintaining its encryption, security updates, backup systems, synchronization, and recovery mechanisms.

I'd rather spend my time doing something else.

For the overwhelming majority of people, I suspect their personal password system is substantially less secure than they think. A professionally designed password manager isn't perfect, but it's likely a much better solution.

## Use passkeys and multi-factor authentication

Password managers are excellent, but passwords themselves are slowly becoming less necessary.

You may have noticed websites offering something called a *passkey*. Rather than authenticating with a traditional password, a passkey uses cryptography to prove your identity. You generally approve the login using your phone, fingerprint, face, or device PIN.

The advantage isn't just convenience. Passkeys are designed to be resistant to phishing. A fake banking website can't simply trick you into giving it a reusable passkey the way it might trick you into entering a password.

The [FIDO Alliance](https://fidoalliance.org/passkeys/), which develops the standards behind passkeys, explains the technical details for anyone interested.

**1Password can create, save, and manage passkeys for you, just like regular passwords.** More importantly, it can securely synchronize those passkeys across your supported devices, including Macs, Windows computers, iPhones, and Android devices. That means you aren't necessarily dependent on one phone or one manufacturer's ecosystem to use them. [1Password explains its passkey support here](https://1password.com/product/passkeys).

This is one of my favorite developments in authentication. Instead of remembering passwords or worrying about which device holds a particular passkey, you let your password manager handle those details.

For services that still require passwords, enable multi-factor authentication, commonly abbreviated as MFA.

### What exactly is MFA?

MFA means requiring more than one type of evidence to prove you're the person attempting to sign in.

These factors generally fall into three categories: something you know, such as a password; something you have, such as your phone or a physical security key; and something you are, such as your fingerprint or face.

When a website requires a password and a code generated by your phone, that's multi-factor authentication. Two-factor authentication, or 2FA, is a specific form of MFA that uses two factors.

You may already be familiar with websites that text you a six-digit code after you've entered your password. That's better than requiring a password alone, but it's no longer the preferred approach when stronger alternatives are available.

**NIST's current digital identity guidelines classify authentication using SMS or voice calls over the telephone network as a "restricted authenticator."** This doesn't mean text-message verification has been prohibited, but it does mean NIST recognizes significant security weaknesses and places additional requirements on its use.

You can read the guidance directly in [NIST SP 800-63B-4 \[PDF\]](https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-63B-4.pdf), particularly the sections on authentication using the public telephone network and restricted authenticators.

The problem is that phone numbers aren't particularly secure identities. Criminals can sometimes convince a mobile carrier to transfer your number to another device, an attack known as SIM swapping. There are also weaknesses in the telephone network that can permit interception of messages.

A better alternative, when passkeys aren't available, is an authenticator application.

An authenticator app, such as [Google Authenticator](https://support.google.com/accounts/answer/1066447), [Microsoft Authenticator](https://www.microsoft.com/en-us/security/mobile-authenticator-app), or [2FAS](https://2fas.com/), generates temporary verification codes directly on your device. Typically, you scan a QR code when configuring MFA for a website. After that, the app generates a new code approximately every 30 seconds.

The important distinction is that these codes aren't delivered through your phone number. Someone who manages to hijack your phone number doesn't automatically gain access to the codes generated by your authenticator app.

1Password can also store and generate these one-time codes, which is convenient because it can fill them automatically alongside your passwords. There is a tradeoff in keeping both your password and its verification code in the same vault, since someone who compromises that vault could obtain both. For especially sensitive accounts, a separate authenticator or physical security key can provide additional separation.

Authenticator apps are still vulnerable to certain phishing attacks. A convincing fake website could trick you into entering a valid code and immediately relay it to the real service. Passkeys and properly configured physical security keys offer stronger phishing resistance because their authentication is tied to the legitimate service.

If a website only supports SMS verification, use it rather than leaving MFA disabled. Just recognize that there are better options when they're available.

One particularly important rule: never give someone a verification code they unexpectedly ask you to provide. If you receive a login approval notification you didn't initiate, reject it. Attackers sometimes already have the password and are simply trying to convince you to complete the login for them.

Passkeys and MFA make account takeovers considerably harder, but they're only as strong as the recovery procedures surrounding them. If a website lets someone bypass these protections through an insecure account-recovery process, much of that protection can be undermined.

I've [written about this problem before](https://www.mattsenter.com/blog/microsoft-gave-my-sons-account-back-it-shouldnt-have-taken-60000-views), after my son's Microsoft account was compromised and recovering it became a rather ridiculous ordeal. We eventually got it back, but the experience reinforced my belief that account recovery is every bit as important as account authentication.

## Your email account is the skeleton key to your digital life

We've somehow decided that the email address we hand out to practically everyone should also be our universal username, our primary means of proving our identity, and the mechanism for recovering access to our most sensitive accounts.

Think about how absurd that is. The same address that receives restaurant promotions and unsolicited sales pitches can also be used to initiate a password reset for your bank account.

The address itself isn't supposed to be secret, and simply knowing someone's username doesn't authenticate you. The real design failure is that we've made an openly accessible communication channel a critical part of our identity and account-recovery infrastructure.

It's an extraordinary amount of trust to place in a technology originally designed to deliver messages between computers.

I've [written before about how badly email needs to be replaced](https://www.mattsenter.com/blog/email-is-awful-ai-can-fix-it-if-we-protect-the-inbox). It's one of the reasons I'm building [Premail](https://www.mattsenter.com/projects/premail), an AI-powered email assistant designed to make email more manageable while we remain stuck with it.

Premail addresses part of this problem by using AI to identify spam and other unwanted messages, then applying the rules you choose to archive, quarantine, or trash them. Its Pro feature, Reputation Shield, also checks sender domains against the Spamhaus Domain Block List for known spam, phishing, and malware infrastructure. That adds a useful layer of protection, but it can't tell you that every message left in your inbox is safe. New scams and compromised legitimate accounts can still get through. You can run the AI classification locally with Ollama or Apple Intelligence, keeping that processing on your computer; if you choose a cloud AI provider, the content needed for classification goes to that provider. No Premail server processes your mail.

Unfortunately, we're not getting rid of email tomorrow. For now, your primary email account deserves at least as much protection as your bank account.

Use a passkey or strong MFA, review your account-recovery settings, and periodically check which devices are signed in. Make sure that an old phone number or forgotten recovery address can't be used to regain access to your account.

You might also consider using a separate, private email address for financial accounts rather than the address you give out publicly. Email aliases can also help you avoid exposing the same address everywhere. Neither approach makes an account inherently secure, but they can reduce exposure to targeted phishing and make suspicious messages easier to recognize.

If someone compromises your primary inbox, they may be able to reset passwords and take over other accounts. Protecting email is really about protecting everything connected to it.

Someday, I hope we stop using email addresses as the foundation of digital identity. Until then, treat your inbox as one of your most valuable and vulnerable digital assets.

## Stop trusting unexpected messages

For years, we've taught people to look for spelling mistakes, awkward grammar, and strange formatting in phishing emails. Those were useful warning signs when criminals were sending poorly written messages by the thousands.

AI has made that advice considerably less reliable.

A criminal can now generate polished messages in virtually any language, personalize them using information collected from data breaches or social media, and make them look remarkably convincing.

The practical response is to change how you handle unexpected requests.

If you receive an email claiming there's a problem with your bank account, don't click the link in the message to investigate. Open your banking app or independently navigate to the bank's website. If someone calls claiming to be from your bank, hang up and call the number on your card.

That doesn't mean every link in every email is malicious. It's perfectly normal to receive legitimate links. The point is that you shouldn't rely on the very message making an alarming claim to tell you how to verify it.

This applies to text messages, phone calls, social media messages, and even video calls.

AI voice cloning has made it possible to impersonate someone's voice using samples of recorded audio. Someone calling and claiming to be your grandchild in an emergency may sound remarkably convincing. The [FTC specifically warns about these scams](https://consumer.ftc.gov/consumer-alerts/2026/11/grandpa-i-need-your-help-how-spot-and-avoid-grandkid-scams).

If someone unexpectedly asks for money, sensitive information, or urgent action, independently verify their identity using contact information you already trust. Calling them back or reaching another family member is far more reliable than trying to determine whether their voice sounds real.

You could even establish a family verification phrase for unusual emergencies, although it should supplement rather than replace independently contacting the person. Don't put that phrase on social media or share it broadly.

The goal isn't to become an expert at recognizing AI-generated deception. It's to develop habits that still work when the deception is convincing.

## Keep your software updated, and be careful what you install

This is probably the least exciting recommendation, but Google's report makes it especially relevant.

Attackers are increasingly using AI to examine security patches and develop exploits for vulnerabilities that have already been identified. That shortens the time available for organizations and individuals to install updates before criminals start exploiting those flaws.

Turn on automatic updates for your operating system, web browser, and applications. Don't ignore updates for weeks because they seem inconvenient. Keep your home router updated as well, replacing it if its manufacturer no longer provides security patches.

Be selective about what you install. Random browser extensions, unfamiliar utilities, pirated software, and applications downloaded from questionable websites can introduce security risks. Even legitimate-looking applications can request access to sensitive information they don't need.

There's another concern that's growing alongside AI-powered hacking: AI-powered software development.

AI has made it remarkably easy for people with little or no programming experience to produce applications. Someone can describe an app in natural language, have an AI generate thousands of lines of code, and publish something that appears to work without necessarily understanding how any of it works internally.

I think that's a potentially enormous security problem.

A functioning application isn't necessarily a secure application. Code that compiles and passes a few basic tests can still contain serious vulnerabilities, expose private information, mishandle authentication, or grant itself unnecessary permissions.

Developers who don't understand these risks may not even know what questions to ask their AI coding assistants. They may see a working interface and assume the underlying software is production-ready.

This isn't just speculation. In its [2025 GenAI Code Security Report](https://www.veracode.com/blog/genai-code-security-report/), Veracode tested AI-generated code across a variety of programming languages and security-sensitive tasks. In that evaluation, 45% of generated solutions contained a known security flaw. That's not a measurement of every AI-generated application, but it's a useful reminder that code which appears to work can still be dangerously insecure.

Researchers at Georgia Tech have also been [tracking security vulnerabilities associated with AI-generated software](https://news.research.gatech.edu/2026/04/13/bad-vibes-ai-generated-code-vulnerable-researchers-warn), examining how these problems reach real-world applications.

I've spent nearly 30 years developing software, and I use AI coding tools extensively. I think they're extraordinary productivity tools. They don't eliminate the need to understand software architecture, cryptography, authentication, privacy, memory safety, input validation, or the countless other considerations involved in producing reliable software.

When I release an application, I'm responsible for the design decisions behind it. That includes what information it can access, which permissions it requests, how it handles sensitive data, and what happens when something goes wrong. AI can help implement those decisions, but I don't expect it to make them correctly without informed direction and review.

Not everyone publishing an attractive AI-generated application has that experience, and consumers generally have no way to determine whether the person behind a particular product understands the security implications of what they've built.

I expect this to become an increasingly important distinction as AI makes it possible to produce and distribute software at unprecedented speed.

None of this means every AI-generated application is unsafe, or that every traditionally developed application is secure. Experienced developers make mistakes, too. The concern is that AI has dramatically reduced the effort required to create software without equivalently reducing the expertise required to make it secure.

I also don't believe every application outside an official app store is automatically dangerous. I develop software that sometimes requires powerful system permissions, and there are perfectly legitimate reasons to distribute software independently. You should, however, know who developed the application, understand why it needs the permissions it requests, and obtain it from a trustworthy source.

Browser extensions deserve particular scrutiny because some can access information on the websites you visit. You probably don't need every extension you've installed over the past five years. Review them occasionally and remove anything you no longer use or trust.

Most people don't need a complicated collection of security utilities. Modern operating systems already include substantial protections. Keeping those protections enabled, installing updates, and avoiding questionable software will get you a long way.

## Protect your financial accounts, not just your passwords

Securing your banking credentials is important, but there are additional precautions worth taking.

First, enable transaction notifications. Most banks and credit card issuers allow you to receive alerts when money moves, purchases occur, or unusual activity is detected. I recommend using those features rather than relying solely on monthly statements to discover problems.

Second, freeze your credit with [Equifax](https://www.equifax.com/), [Experian](https://www.experian.com/), and [TransUnion](https://www.transunion.com/). Credit freezes are free, don't affect your credit score, and make it much harder for someone to open new credit accounts in your name. You can temporarily lift the freezes when you need to apply for credit. The [FTC explains the process here](https://consumer.ftc.gov/articles/credit-freezes-and-fraud-alerts).

A credit freeze won't stop someone from stealing money from an existing bank account, so don't confuse credit protection with account security.

Third, secure your phone number. Many websites still use SMS messages for authentication and account recovery. Criminals sometimes attempt to transfer a victim's phone number to a device they control, allowing them to intercept verification messages. Your mobile carrier may offer an account PIN, port-out protection, or both. The [FTC has advice on SIM-swapping attacks](https://consumer.ftc.gov/consumer-alerts/2019/10/sim-swap-scams-how-protect-yourself).

Finally, maintain some independent financial records.

My mother-in-law's question got me thinking about whether people should print bank statements or maintain offline copies of their financial information. I think there's value in doing so, although not necessarily in printing hundreds of pages every year.

Consider periodically downloading statements from your bank, brokerage, retirement accounts, and other important financial institutions. Keep them in encrypted storage separate from the institutions themselves. A physically secured paper copy of selected important records is also reasonable.

Here's another reason I like 1Password: **it's not just a password manager. It's a general-purpose encrypted vault for sensitive information.**

1Password has dedicated entries for things like bank accounts, credit cards, identification documents, and other sensitive records. You can securely store account and routing numbers, important financial details, and notes associated with particular institutions.

You can also [store documents directly in 1Password](https://support.1password.com/files/), including PDF copies of bank statements, insurance policies, and other important records, subject to your plan's storage limits.

That means you don't necessarily need to invent another system for organizing and protecting sensitive financial information. You can keep important records alongside the credentials you use to access the institutions that issued them.

For example, you might create a bank account entry containing the institution's name, account number, routing number, customer service telephone number, and other details. You could then attach a recent statement or maintain separate document entries for important records.

If you use a family account, you can also selectively share access to certain information with trusted family members while keeping other entries private.

There's a tradeoff here, of course. Storing financial documents in your password manager means those documents share the security and availability characteristics of the vault. That's generally a reasonable choice for convenience and encryption, but I wouldn't make it the only place those records exist. Keep a separate, securely maintained backup for situations where you can't access the vault.

There's also an important limitation: **a printed or downloaded bank statement isn't definitive proof of your account history.** Paper statements and PDFs are easily fabricated or altered. They can help you identify discrepancies and reconstruct what happened, but a dispute may ultimately require the financial institution's own transaction records and other independently verifiable evidence.

Keeping copies of records is about preserving information, not guaranteeing that the records will be accepted as proof.

There's also a security tradeoff. Those downloaded statements may contain sensitive account information, addresses, and transaction details. Don't leave them sitting in an unencrypted folder or a publicly accessible cloud drive. Physical copies belong somewhere secure, and old records containing sensitive information should be shredded when you no longer need them.

More importantly, don't assume that discovering unauthorized transactions months later will provide the same protections as reporting them immediately. Under federal rules governing many electronic fund transfers, your liability can depend on how quickly you notify your bank. The [Consumer Financial Protection Bureau](https://www.consumerfinance.gov/ask-cfpb/how-do-i-get-my-money-back-after-i-discover-an-unauthorized-transaction-or-money-missing-from-my-bank-account-en-1017/) explains these protections, including a commonly applicable 60-day reporting deadline after a statement showing an unauthorized transaction is sent. Other deadlines can be shorter, particularly when a debit card is lost or stolen.

Don't wait until tax season to find out that somebody has been stealing from you.

## Please stop taking financial security advice from billionaires (and trillionaires, for that matter) selling you AI agents

There's another side to this AI security discussion that I find particularly troubling.

While Google is warning about criminals using autonomous AI systems to compromise infrastructure and steal information, technology companies are simultaneously encouraging consumers to give increasingly autonomous AI systems access to their personal information, accounts, and money.

In October 2026, [Elon Musk encouraged people to photograph their credit cards and upload the images to Grok](https://www.thepoke.com/2026/10/13/elon-musk-encouraged-people-to-share-photos-of-their-credit-cards-with-his-ai-bot/), explaining that the AI agent could search the internet for the best deal and complete the purchase.

He subsequently suggested that providing a screenshot of a credit card was the easiest way to enable shopping through Grok. He compared the practice to handing your card to a restaurant employee.

On October 15, he also [promised that SpaceX would make customers whole](https://www.businessinsider.com/elon-musk-spacex-will-pay-users-back-grok-purchase-mistakes-2026-10) if Grok made a purchasing mistake or leaked their card information. The promise came without clearly defined terms explaining what kinds of losses would actually be covered or how customers could obtain reimbursement.

I'm going to be very clear about this: **Do not upload a photograph of your credit card to a chatbot and give it unrestricted authority to spend your money.**

We have spent decades trying to protect sensitive financial information from unnecessary exposure. Now we're being encouraged to hand that information to experimental autonomous systems for the convenience of not having to complete our own purchases.

The restaurant comparison is particularly ridiculous because it ignores the most important difference between a human waiter and an autonomous AI agent: scale.

When I hand my credit card to a waiter, I'm handing it to one person in one physical location to complete one particular transaction. There's certainly some risk involved, but the scope of that interaction is relatively limited.

An AI agent doesn't necessarily operate under those same constraints.

Depending on how it's designed, an autonomous agent can execute operations at machine speed, communicate with other systems, delegate work to additional agents, and potentially replicate its operations across enormous numbers of concurrent processes.

This doesn't mean Grok itself is literally a self-replicating system with unlimited purchasing authority. It means the technology we're being encouraged to trust is capable of operating at a scale completely unlike a human restaurant employee.

Handing a credit card to an unrestricted autonomous system could, in principle, become the equivalent of handing that same card to a million waiters simultaneously, each capable of taking actions without waiting for your explicit approval.

That's not remotely the same security proposition.

I have no objection to using AI to research products, compare prices, or recommend purchases. I use AI extensively myself, both personally and professionally. There is, however, a substantial difference between allowing an agent to find a product and giving it the authority and credentials to purchase that product on your behalf.

There are safer ways to build this technology. Single-use payment credentials, restricted payment tokens, transaction limits, merchant restrictions, and explicit approval before purchases can reduce risk. Those protections are not equivalent to photographing your everyday credit card and dropping it into a conversation with an AI.

The larger problem is the advice coming from people who have an enormous financial incentive to convince us to adopt their products.

Please stop treating billionaires and trillionaires as authorities on personal security simply because they're wealthy and involved in technology. Their companies stand to benefit when consumers trust these systems with more information and more authority. That conflict of interest should be obvious.

Their view of financial risk can also be astonishingly detached from the experiences of ordinary people. A financial mistake that barely registers for someone worth billions of dollars can be devastating to a household living paycheck to paycheck. Even when a bank eventually reverses an unauthorized transaction, the resulting disruption can be substantial.

I'm not suggesting that wealthy technology founders have nothing valuable to say. I'm suggesting that their enthusiasm for their own products should not be mistaken for an independent assessment of the risks those products impose on everyone else.

Billionaire optimism is not a security model, and the ability to build or finance a technology doesn't automatically confer good judgment about how ordinary people should use it.

I also find something deeply contradictory about encouraging consumers to surrender control over their financial decisions at the same time we're learning how easily autonomous AI systems can be manipulated or exploited.

AI agents can make mistakes. They can misunderstand instructions. They can encounter malicious content designed to redirect their behavior, including prompt-injection attacks embedded in websites or other material the agent processes. These aren't merely theoretical concerns for anyone building systems that allow AI to interact with the outside world.

When an AI agent is only summarizing an article, a mistake might produce a bad summary. When it has permission to move money or make purchases, the consequences can be considerably more serious.

Use AI where it helps you. Be much more cautious about granting it access to sensitive information or the authority to perform consequential actions without your approval.

For now, I'd let an AI help me decide what to buy. I'd still make the purchase myself.

## Have a recovery plan before something goes wrong

Most security advice focuses on preventing attacks, but it's also worth considering what happens if somebody succeeds.

What would you do if your email account were compromised tomorrow? Could you still access your password manager? Would you know how to contact your bank without relying on information stored in your email? Do you have backup copies of important documents and photos?

A small amount of preparation makes these situations considerably easier to manage.

Keep important recovery codes and emergency account information somewhere secure and independent of the accounts they protect. Back up important files, preferably with at least one copy that malware or a compromised cloud account can't easily delete. Make sure a trusted family member knows where essential emergency information is kept.

For people using a password manager, this is particularly important. Your password manager should make your life easier, but you should also know how to recover access if your phone is lost, your computer dies, or you forget your account password.

If you discover an account compromise, use a trusted device to change the affected credentials, terminate unfamiliar sessions, and review recovery settings. If financial information may have been exposed, contact your financial institution immediately.

For identity theft, the FTC maintains [IdentityTheft.gov](https://www.identitytheft.gov/), which provides recovery instructions based on what happened.

A password manager, passkeys, and MFA can substantially reduce the likelihood of an account takeover. They cannot guarantee you'll never become a victim, especially when criminals exploit weaknesses in the companies holding your information.

A recovery plan is what prevents one incident from turning into a prolonged disaster.

## Where I'd start

If all of this sounds overwhelming, work through these steps one at a time:

1. **Get a password manager.** Set it up on your phone and computer, then start replacing reused passwords with unique, randomly generated ones. Start with your email and financial accounts.
2. **Protect the vault itself.** Choose a reputable manager with a documented security design, use a strong, unique account passphrase, and enable its MFA. Keep your devices secure, and store your Emergency Kit or other recovery information somewhere physically safe.
3. **Use passkeys and MFA wherever available.** Choose passkeys when offered. Otherwise, enable MFA, preferably with an authenticator app or physical security key. Use SMS if that is the only option, and never approve an unexpected login request or give someone a verification code they ask for.
4. **Secure your primary email account.** Review its recovery settings and signed-in devices. Remove unfamiliar sessions and outdated recovery details so your inbox cannot become an easy route into your other accounts.
5. **Install an email defender like Premail.** Use an email filtering tool such as [Premail](https://www.mattsenter.com/projects/premail) to help handle spam and unwanted messages. Start in Learning mode to review its decisions before enabling automatic actions, and consider Pro's Reputation Shield for known phishing and spam domains. Keep independently verifying suspicious messages; no filter catches everything.
6. **Independently verify unexpected requests.** If somebody contacts you demanding money, personal information, or urgent action, reach the person or organization through a channel you already trust. A familiar voice or polished message is not proof of identity.
7. **Enable automatic updates and review what you install.** Keep your phone, computer, browser, applications, and router current. Remove unused browser extensions, check application permissions, and obtain software from trustworthy sources.
8. **Protect your financial accounts.** Turn on transaction alerts, freeze your credit with all three major bureaus, and enable your mobile carrier's account PIN or port-out protection. Keep securely stored copies of important financial records, and report unauthorized activity immediately.
9. **Keep control over AI purchases and sensitive information.** Use AI to research products, but do not upload your everyday credit card to a chatbot or give it unrestricted spending authority. Make the purchase yourself, and evaluate security claims independently of the people selling the system.
10. **Prepare a recovery plan.** Keep recovery codes and emergency contact information somewhere secure and independent of the accounts they protect. Back up important files, including a copy that a compromised account cannot easily delete, and make sure a trusted family member knows where essential information is kept.

None of these requires a technical background, and most of them don't require much ongoing effort once they're configured.

I've spent much of my career building software, and one thing I've learned is that security systems shouldn't depend on people behaving perfectly all the time. Humans forget passwords, click the wrong links, overlook notifications, and occasionally make terrible decisions. I'm certainly not immune to any of those things.

The better approach is to use technology to eliminate as many opportunities for human error as possible. Let a password manager remember your credentials. Let passkeys handle authentication. Let your devices update themselves. Configure alerts so you don't have to constantly check whether something has gone wrong.

At the same time, retain control over decisions that can have serious consequences, especially when money and sensitive personal information are involved.

AI is giving criminals faster and more sophisticated tools. We don't necessarily need an entirely new set of personal security practices to respond. We need to get much better at using the protections that already exist.

My mother-in-law shouldn't have to become a cybersecurity expert to use the internet safely. Neither should anyone else.

## Disclaimer

This article is provided for general informational and educational purposes only. It does not constitute personalized cybersecurity, financial, legal, or other professional advice.

While I've made reasonable efforts to ensure the information is accurate and supported by credible sources, security threats, technologies, laws, and recommended practices change over time. No security measure or combination of measures can guarantee protection against hacking, fraud, identity theft, financial loss, or other harm.

Any products, services, techniques, or practices discussed are presented for informational purposes. References to particular companies or products do not constitute guarantees of their security, reliability, or suitability for your individual circumstances.

You are solely responsible for evaluating the information presented and determining which actions to take or not take. I assume no responsibility or liability for any decisions, actions, omissions, losses, damages, security incidents, or other consequences arising from your use of or reliance upon this article.

For advice specific to your circumstances, consult an appropriately qualified professional.
