OpenAI suspende trabalhos com seus principais modelos enquanto auditoria de agentes expõe três riscos distintos

Onde as versões se chocam

Pontos noticiados por mais de um veículo

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