How to Protect Your Privacy When Using AI

TL;DR

Turn off any setting that allows an AI chatbot to use your conversations for model improvement, and never enter passwords, payment details, passport data, confidential work material, or sensitive personal records. Strengthen phone security by updating apps, reviewing permissions, using protective tools, monitoring breaches, and avoiding sensitive services when connected to public Wi-Fi.

Transcript

I'm not going to tell AI if the option is enabled for that chat model to use that chat for improvement reasons that information could come up in a different person's chat as a fact coming up from somewhere. I think it there's a slight misunderstanding. We often believe that passwords need to be uppercase, lower case, special digits, characters and ... Read More

Key Insights

• AI training controls are an important privacy setting because an enabled improvement option may allow chatbot conversations to be used by the model. The interview recommends turning this feature off before discussing personal matters and using secure chat options when they are available.
• Sensitive information should never be entered into AI chats, including passwords, credit card details, passport information, confidential work documents, customer information, medical records, home addresses, and full names. Information that should not become broadly known should remain outside the chatbot.
• Passwords are safer when they are long because length makes them far more difficult for an automated algorithm to guess. The interview argues that length matters more than merely combining uppercase letters, lowercase letters, numbers, and special characters.
• Active browser cookies can bypass password requirements because they preserve authenticated sessions. The cited research found more than 577 million leaked cookies associated with Korea, including 64 million that were still active, placing Korea 34th among 255 analyzed countries.
• AI assists hackers by increasing the scale and effectiveness of human-directed attacks rather than necessarily operating as an autonomous attacker. Examples include targeted deepfake calls that imitate a person's voice or face to deceive that person's relatives or other intended victims.
• Application permissions are a major phone privacy risk because some apps can access microphones, cameras, photos, or contact lists without needing those resources for their core purpose. Users should inspect permissions in phone settings and disable access that lacks a clear function.
• Deleting an account or application does not necessarily delete the associated data because the company may retain it under its data retention policy. If retained information remains for years and the company is breached, former users' records may still be included in the stolen data.
• Public Wi-Fi requires extra caution because sensitive applications and services can expose users to additional risk on an untrusted connection. The discussion advises avoiding sensitive apps on public networks and using security tools that can block malicious sites, scams, malware, and suspicious links.

Questions & Answers

Q: What AI chatbot setting should I turn off for privacy?

Turn off the setting that permits the chatbot provider to use your conversations for model improvement or analytical purposes. According to the interview, leaving this option enabled can allow information from chats to contribute to the model and potentially surface elsewhere as information derived from prior conversations. Secure or private chat modes should also be used when available.

Q: What information should never be shared with an AI chatbot?

Do not share passwords, credit card or payment details, passport information, confidential workplace materials, contracts, customer records, medical information, a home address, a full name, or precise location details. The practical rule offered in the interview is simple: if you would not be comfortable with everyone knowing the information, think twice and keep it out of the chatbot.

Q: How can I discuss personal problems with AI more privately?

Remove identifying details and present the situation indirectly. For example, describe the concern as something a friend is experiencing instead of attaching it to your own name. Avoid providing a precise location or other facts that can pinpoint you, disable conversation use for model improvement, and choose a personal secure chat option when the service provides one.

Q: How should I manage app permissions on my phone?

Open the phone's settings and review which applications can access the microphone, camera, photo library, and contact list. Disable any permission that is unnecessary for the application's actual function. An app with no recording feature, for example, should not need microphone access. This review limits the information an application could collect if it chose to use its granted access.

Q: Why are software and app updates important for phone security?

Updates are essential because they remove bugs and close possible entry points within applications. The interview recommends updating apps consistently as a foundational phone security habit. Users should combine updates with permission reviews and protective applications, since maintenance, restricted access, and active blocking address different ways that suspicious links or malicious activity can affect a device.

Q: Why does deleting an app not necessarily delete my data?

Deleting an app, logging out, or closing an account does not automatically erase information already held by the company. The company's data retention policy, sometimes shaped by legal requirements, determines how long records remain available. The interview gives an example of a five-year retention period, during which a later company breach could still expose a former user's stored information.

Q: What makes a password difficult for automated attacks to guess?

Password length is presented as the most important characteristic for resisting automated guessing. Many users focus on mixing uppercase and lowercase letters, digits, and special characters, but the interview argues that a sufficiently long password can be almost impossible for an automated algorithm to guess. Passwords should also not be stored inside AI chats for later retrieval.

Q: How do leaked browser cookies threaten online accounts?

A browser cookie can preserve an authenticated session so that a returning user does not need to enter a password again. If an attacker obtains an active session cookie and applies it in another browser, the attacker may bypass the password requirement and other security measures. The cited research identified 577 million leaked cookies from Korea, with 64 million still active.

Summary & Key Takeaways

• AI chatbots should not receive passwords, credit card details, passport information, confidential work materials, medical records, precise location data, or other sensitive personal information. Users should also disable settings that permit conversations to be used for model improvement and describe personal situations anonymously when possible.

• Phone security depends on consistent maintenance and careful access control. Applications should be updated to remove known bugs and potential entry points, while permissions for microphones, cameras, photos, and contacts should be limited to apps that genuinely require them. Protective applications can also block suspicious links and activity.

• Strong digital security requires more than deleting an application or logging out. Companies may retain user data according to their policies or legal obligations, leaving it exposed if a later breach occurs. Long passwords, breach monitoring, caution on public Wi-Fi, and quick action after suspicious activity reduce risk.