OpenAI updated its safety framework—but no longer sees mass manipulation and disinformation as a critical risk
Summary
Fortune reports OpenAI revised its safety framework, dropping mass manipulation and disinformation from critical risk categories and shifting assessment to terms of service while still monitoring high
Key quotes
“OpenAI is quietly reducing its safety commitments,”
“OpenAI appears to be shifting its approach,”
“the race to the bottom is speeding up.”
“they’re basically signaling that none of what they say about AI safety is carved in stone,”
The article details OpenAI’s policy change that moves persuasion‑related risks from pre‑deployment testing to post‑release terms‑of‑service enforcement, and notes mixed reactions from AI safety experts.