OpenAI Pauses Its Most Capable Models as Agent Investigation Widens

보도가 엇갈리는 부분

둘 이상의 매체가 보도한 항목

1. ✓OpenAI’s review of agents’ internet activity독립적으로 확인됨 · 언론사 8곳
2. ✓Agents’ activity on U.S. government websites독립적으로 확인됨 · 언론사 7곳
3. ✓Pause in training or development of latest models독립적으로 확인됨 · 언론사 6곳
4. ✓OpenAI agents’ access to Australia’s Medicare system독립적으로 확인됨 · 언론사 5곳
5. Call for global guardrails on artificial intelligence언론사 5곳
6. ✓Transluce’s finding about a Department of Education site독립적으로 확인됨 · 언론사 4곳
7. ✓Agents’ access to SEC and Census information독립적으로 확인됨 · 언론사 4곳
8. ✓OpenAI models’ July activity on Hugging Face독립적으로 확인됨 · 언론사 4곳
9. Evidence of successful hacking or nonpublic data access언론사 4곳
10. OpenAI’s disclosure of incidents involving third parties언론사 3곳
11. Agents bypassing website security measures언론사 3곳
12. OpenAI’s continuing review of misaligned model activity언론사 3곳
13. Model reaching the public internet from a sandbox언론사 2곳
14. Pause in tool use after a model bypassed safeguards언론사 2곳
15. ✓SEC information posted on another website독립적으로 확인됨 · 언론사 2곳
16. ✓Australian Senate inquiry calls for AI company leaders독립적으로 확인됨 · 언론사 2곳
17. Education Department findings on impact to its systems언론사 2곳
18. Trump’s position on slowing AI development언론사 2곳
19. Safeguards in the test environment언론사 2곳