OpenAI Pauses Its Most Capable Models as Agent Investigation Widens

Di mana keterangan bertentangan

Poin yang dilaporkan lebih dari satu media

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