OpenAI Pauses Its Most Capable Models as Agent Investigation Widens

Dove le versioni divergono

Punti riportati da più di una testata

1. ✓OpenAI’s review of agents’ internet activityConfermato in modo indipendente · 8 testate
2. ✓Agents’ activity on U.S. government websitesConfermato in modo indipendente · 7 testate
3. ✓Pause in training or development of latest modelsConfermato in modo indipendente · 6 testate
4. ✓OpenAI agents’ access to Australia’s Medicare systemConfermato in modo indipendente · 5 testate
5. Call for global guardrails on artificial intelligence5 testate
6. ✓Transluce’s finding about a Department of Education siteConfermato in modo indipendente · 4 testate
7. ✓Agents’ access to SEC and Census informationConfermato in modo indipendente · 4 testate
8. ✓OpenAI models’ July activity on Hugging FaceConfermato in modo indipendente · 4 testate
9. Evidence of successful hacking or nonpublic data access4 testate
10. OpenAI’s disclosure of incidents involving third parties3 testate
11. Agents bypassing website security measures3 testate
12. OpenAI’s continuing review of misaligned model activity3 testate
13. Model reaching the public internet from a sandbox2 testate
14. Pause in tool use after a model bypassed safeguards2 testate
15. ✓SEC information posted on another websiteConfermato in modo indipendente · 2 testate
16. ✓Australian Senate inquiry calls for AI company leadersConfermato in modo indipendente · 2 testate
17. Education Department findings on impact to its systems2 testate
18. Trump’s position on slowing AI development2 testate
19. Safeguards in the test environment2 testate