1406 articles tagged with Safety/Alignment
🤖 AI safety debate shifts toward resilience over regulationPolicymakers are increasingly focusing on improving AI resilience rather than implementing extraordinary government inter...
Feds freaked over Fable 5 after simple 'fix this code' prompt, not jailbreak
🤖 Feds freaked over Fable 5 after simple 'fix this code' prompt, not jailb...📝 The “jailbreak”...https://www.theregister.com/security/2026/06/15/feds-freaked-over-fable-5-after-sim...
A practical comparison of RLHF, DPO, IPO, and KTO — what each method actually does under the hood, how their data and compute requirements differ, and when to pick one over the oth...
Feds freaked over Fable 5 after simple 'fix this code' prompt, not jailbreak, says researcherAccording to the one person who actually read the research paperhttps://www.theregister...
🤖 New nonprofit aims to accelerate AI alignment research as superintelligence loomsResearchers are forming new nonprofit organizations to accelerate AI alignment research because c...
The US government's Anthropic models ban was never about an AI jailbreak | TechCrunchThe Trump administration's decision that forced Anthropic to pull its latest cybersecurity mode...
Anthropic's fable got shut down today after a jailbreak exposed its 120,000-character system prompt — now public on GitHubThe industry response: don't depend on one vendor. Self-ho...
Artificial intelligence firm Anthropic is working to restore access to its latest models after a White House directive forced the company to pull them down.
Anthropic says the jailbreak triggering the Commerce Department order is narrow and that OpenAI's GPT-5.5 produces the same output. Amazon's CEO reportedly flagged the vulnerabilit...
Alignment-research scaffold (autoresearch-style) for LLM guardrails: search over a single policy.md surface
Anthropic refused to fix Fable 5 jailbreak - AI news #Shorts David Sacks said the US government warned Anthropic that Claude ...
Import AI 461: "Alignment is not on track"; FrontierCode; and synthetic research internshttps://importai.substack.com/p/import-ai-461-alignment-is-not-on#AI #Research #Ethics
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe no...
A look at the structural improvements in version 1.9.0 — and why an MIT-licensed red teaming...
Entrepreneurs have heard for years that artificial intelligence will change how companies operate. In many ways, that prediction has already come true. AI can draft emails, summari...
司南：个性化 AI 任务总控 Skills 系统 /COMPASS: Personal Alignment Skills OS for AI Agents
Singapore's Infocomm Media Development Authority and Microsoft signed a memorandum of understanding to strengthen their collaboration on artificial intelligence safety and security...
We introduce TuneJury, an open, instance-level pairwise reward model for text-to-music that predicts a music preference score from a text prompt and an audio clip. The released che...
Fable jailbreak?: Swift shutdown of Anthropic's new LLM may have been based on Amazon researchers figuring out how to circumvent safeguardshttps://www.theverge.com/ai-artificial-in...
The Trump administration restricted access to Anthropic's Mythos 5 AI modEl after Amazon researchers jailbroke it, citing national security concerns. The move has sparked debate ov...
'It's not a jailbreak' — Research leading to U.S. export restrictions on top Anthropic models was for defense, cybersecurity CEO says | Fortune"It was Defense Oriented Prompting (D...
A new CVE against Cursor, a LiteLLM supply-chain backdoor, and a study showing image-only injection...
Anthropic says all AI models can be hacked."We suspect that perfect jailbreak resistance is not currently possible for any model provider." "...it is likely that universal jailbrea...