Old version
May 12, 2026 06:15 UTC
e6aff17aaa45ede5b6c9ae5aa7145d80dd3d6023da3e4f89239abbae35c948a6
CA-V-002508
New version
May 24, 2026 00:02 UTC
228aa71109d71bf3c629fd455c29373a73e8706f1ed68fcdeb212fcac398f524
CA-V-002932
Share 𝕏 Share in Share
Change Summary
OpenAI updated its Safety Standards document on May 24, 2026, reorganizing how it describes its AI safety approach. Three sentences about safety testing and real-world feedback were removed, and the remaining language now emphasizes teaching AI right from wrong, filtering harmful content, conducting internal evaluations with experts, and using real-world feedback. The document is shorter and more tightly focused on four core safety practices, though the substantive safety commitments described remain similar.
low severity
0 Sentences added
3 Sentences removed
3 Sentences modified
11 Sentences before
8 Sentences after
Added
Removed
Modified
BeforeAfter
0Safety & responsibility | OpenAI Safety Safety at every step We believe in AI’s potential to make life better for everyone, which means making it safe for everyone Teach Your browser does not support the video tag.0OpenAI Safety Safety at every step We believe in AI’s potential to make life better for everyone, which means making it safe for everyone Teach We start by teaching our AI right from wrong, filtering harmful content and responding with empathy.
1We start by teaching our AI right from wrong, filtering harmful content and responding with empathy.1Read more Test We conduct internal evaluations and work with experts to test real-world scenarios, enhancing our safeguards.
2Read more Test Your browser does not support the video tag.2Read more Share We use real-world feedback to help make our AI safer and more helpful.
3We conduct internal evaluations and work with experts to test real-world scenarios, enhancing our safeguards.Removed
4Read more Share Your browser does not support the video tag.Removed
5We use real-world feedback to help make our AI safer and more helpful.Removed
Stay ahead of the changes

Watch this before it changes again

Follow unlimited companies, monitor the clauses that matter across every platform, and get the full institutional analysis on what each change obligates you to do.