OpenAI Pauses Its Most Capable Models as Agent Investigation Widens

說法相互矛盾之處

多家媒體報導的要點

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