AI Safety Guide

Everyday safety

Using AI tools without getting hurt: wrong answers, privacy, scams, and over-reliance

Frontier safety

Making sure powerful AI systems stay reliable, controllable, and beneficial

Most AI harm in daily life comes from trusting it too much or sharing too much.

• AI can sound confident and still be wrong
• Check important facts against reliable sources
• Ask for sources, then open them
• Get professional advice for health, legal, and money decisions

• Do not share passwords, ID numbers, or bank details in chats
• Check whether your chats are used for training, and change it if you prefer
• Use work accounts and approved tools for work data
• Delete chats you no longer need

• Voices and videos can be faked
• Verify urgent money requests by calling back on a known number
• Agree a family code word for emergencies
• Be suspicious of pressure, secrecy, and urgency

• Give agents only the access they need
• Keep approvals on for sending, paying, and deleting
• Web pages and emails can hide instructions that try to trick agents, so watch what they do
• Never let an agent handle passwords or one-time codes

• Do not create AI images of real people without consent
• Label realistic AI content when you share it
• Report non-consensual AI images to the platform

• Use age-appropriate settings and parental controls
• Talk about AI companions and over-reliance
• Encourage them to check facts and think for themselves
• Make sure they know they can tell you about anything upsetting

• Follow your company’s AI policy
• Keep confidential data out of unapproved tools
• Review AI work before it reaches customers
• Report AI mistakes so processes improve

• Researchers study how to make AI honest, controllable, and resistant to misuse
• Leading AI companies publish safety frameworks and test models before release
• Governments are creating AI laws, such as the EU AI Act
• Independent testing and transparency are growing

• I check important facts
• I keep sensitive data out of AI chats
• I keep approvals on for agents
• I verify urgent requests another way
• I label AI content I share