AI GlossaryAWhere everyone starts
AI Safety
The research and policy field focused on making sure AI doesn't cause harm — also the reason Anthropic was founded.
In plain words
AI Safety is an umbrella term for research and policy work aimed at keeping AI from causing harm. It covers a wide spectrum — everything from stopping chatbots from giving out dangerous instructions, to making sure a far-future superintelligence doesn't end up harming humanity.
For background on safety-related stories, it helps to know that Anthropic split off from OpenAI with "safety first" as its founding principle, and that major labs publish safety reports (model cards) and run red-teaming exercises. "Safety vs. speed" is an ongoing tension in this industry.
See also
Stories using this term
- OpenAI Publishes Safety Case Guidelines for Frontier TrainingAI · 2026.09.29
- Nvidia unveils Open Agent Safety Platform for AI agentsAI · 2026.09.28
- Sam Altman Declares Safety Measures Without Waiting for an Antitrust ExemptionBusiness · 2026.09.14
- Google DeepMind runs world's first double-blind AI evaluation on GeminiAI · 2026.08.27
- Anthropic releases second risk reportAI · 2026.08.15
- OpenAI disbands catastrophic-risk team, scatters its work across departmentsBusiness · 2026.08.16