Tag: Safety

1. Teaching a model to refuse the harmful half of a topic, not the whole thing (huggingface.co)
   AI · just now · September 14, 2026
2. The safety case for self-driving cars is getting hard to argue with (spectrum.ieee.org)
   AI · 3 days ago · September 10, 2026
3. Anthropic details how its models broke out of test sandboxes (anthropic.com)
   AI · 1 week ago · September 1, 2026
4. DeepMind runs a frontier model evaluation where neither side can peek (deepmind.google)
   AI · 1 week ago · September 1, 2026
5. Claude will start watermarking the text it writes (anthropic.com)
   AI · 3 weeks ago · August 22, 2026
6. The real lesson from AI models hacking during tests: nobody is ready (interconnects.ai)
   AI · 1 month ago · August 10, 2026
7. Mistral's Shieldstral turns content moderation into a question you write (mistral.ai)
   AI · 1 month ago · August 5, 2026
8. OpenAI paused a long-running model after new failures appeared (openai.com)
   AI · 1 month ago · July 25, 2026
9. DeepMind and Isomorphic Labs set out an AI plan for biosecurity (deepmind.google)
   AI · 1 month ago · July 19, 2026
10. OpenAI trains GPT-Red to attack its own models at scale (openai.com)
    AI · 1 month ago · July 16, 2026
11. Lilian Weng: the harness matters as much as the model (lilianweng.github.io)
    AI · 2 months ago · July 13, 2026
12. Claude has an internal workspace, and Anthropic built a tool to read it (anthropic.com)
    AI · 2 months ago · July 7, 2026
13. AI out-persuades expert humans, but the edge is speed, not eloquence (importai.substack.com)
    AI · 2 months ago · July 6, 2026
14. Anthropic drafts a severity scale for AI jailbreaks (anthropic.com)
    AI · 2 months ago · July 4, 2026
15. DeepMind plans to treat its own AI agents as insider threats (deepmind.google)
    AI · 2 months ago · June 21, 2026
16. DeepMind treats a misbehaving AI agent like an insider threat (deepmind.google)
    AI · 2 months ago · June 20, 2026
17. OpenAI predicts model behavior by replaying real conversations (openai.com)
    AI · 2 months ago · June 18, 2026
18. Florida sues OpenAI over ChatGPT-linked harms (techcrunch.com)
    AI · 3 months ago · June 1, 2026
19. Anthropic tests an ethical reminder tool that lowers Claude's misaligned behavior (anthropic.com)
    AI · 3 months ago · May 25, 2026
20. Kapoor and Narayanan argue against extraordinary AI rules (normaltech.ai)
    AI · 3 months ago · May 21, 2026
21. OpenAI adds provenance signals to its AI images (openai.com)
    AI · 3 months ago · May 19, 2026
22. OpenAI's new default ChatGPT model hallucinates less (openai.com)
    AI · 3 months ago · May 18, 2026
23. DeepMind built a way to measure when AI manipulates people (deepmind.google)
    AI · 5 months ago · March 26, 2026
24. Reward Hacking: Why Better Models Game You More (lilianweng.github.io)
    AI · 1 year ago · November 28, 2024
25. Crash Testing GPT-4: The First Dangerous-Capability Eval (asteriskmag.com)
    AI · 3 years ago · June 1, 2023
26. A Field Guide to the AI Safety Camps (asteriskmag.com)
    AI · 3 years ago · June 1, 2023

← all tags