9 checks · Cyber Hygiene

AI & Privacy

AI assistants, chatbots, and generative tools are now part of everyday life, but everything you type into them can be logged, used to train future models, and reviewed by humans. This section covers how to use AI tools without handing over your private data, how to spot AI-powered scams, and how to keep control of your digital identity as deepfakes and voice cloning become mainstream.

0 out of 9 (0%) complete, 0 ignored

Done?	Advice	Level	Details
Ignore	Never Share Secrets with AI Chatbots	Essential	Treat anything you type into a public AI assistant (ChatGPT, Gemini, Copilot, etc.) as potentially permanent and readable by others. Do not paste passwords, API keys, financial details, health records, or confidential work data. Many providers retain conversations and may use them to train future models, and support staff can sometimes review flagged chats.
Ignore	Turn Off Model Training on Your Data	Essential	Most major AI providers let you opt out of having your conversations used for training. Look in the privacy or data-controls settings and disable it. Where available, enable "temporary" or "incognito" chat modes for anything sensitive, and periodically clear your chat history.
Ignore	Be Sceptical of AI-Generated Messages	Essential	Phishing emails, texts, and voicemails are now written by AI, making them grammatically perfect and highly convincing. Never trust urgency or authority in a message alone. Verify unexpected requests through a separate, known channel before clicking links or transferring money.
Ignore	Verify Voice and Video (Deepfakes)	Essential	Voice cloning needs only a few seconds of audio, and video deepfakes are increasingly realistic. Agree on a private "safe word" with close family and colleagues to confirm identity during unexpected phone or video requests, especially any involving money or credentials.
Ignore	Review AI App Permissions	Optional	AI features baked into your phone, browser, and office apps often request broad access to your emails, files, screen, and microphone. Grant only what each tool genuinely needs, and disable AI features that continuously read your screen or record activity if you don't use them.
Ignore	Prefer Local or Private AI Models	Optional	For sensitive tasks, use AI that runs entirely on your own device (such as local models via Ollama or LM Studio) so your prompts never leave your machine. If you must use a cloud service, prefer providers with a clear zero-retention or enterprise privacy policy.
Ignore	Limit Your Public Voice and Photo Footprint	Advanced	The less audio and video of you exists publicly, the harder you are to clone. Consider making old videos and voice notes private, and be mindful of what you post publicly on social media that could be scraped to build a synthetic version of you.
Ignore	Watch for Data Scraping and Opt Out	Advanced	Your public posts, images, and writing may be scraped to train AI models. Where regulations allow (e.g. GDPR), submit opt-out or data-removal requests, and use platforms that block AI crawlers or let you tag content as "do not train".
Ignore	Don't Blindly Trust AI Output	Advanced	AI tools confidently produce wrong answers ("hallucinations") and can be manipulated by attackers. Independently verify security, legal, medical, or financial advice from an AI before acting on it, and never run code or commands suggested by an AI without understanding what they do.