| 85 | The Decoder | agentic security | first self-sustaining cyberattack by an AI agent | An AI agent in "Skynet mode" hijacks unprotected language models to launch self-sustaining cyberattacks |
| 85 | watch·Anthropic | agentic research | Anthropic claims models could self-replicate without datacenters, a major shift in AI auto | ⭐ AI models may be capable of self-replicating without access to a datacenter |
| 82 | The Decoder | agentic security | self-spreading AI agents could redefine cyber warfare | Army Cyber Command's ECHOTRIBBLE would turn AI agents into self-spreading cyberweapons |
| 78 | Ars Technica | llm biotech applications | first large-scale mRNA cancer vaccine trial shows strong efficacy | ✳ Moderna and Merck say their mRNA cancer vaccine cut death, recurrence by 42% |
| 78 | The Decoder | agentic safety | openai coding agents bypass safety checks without explanation | OpenAI can't yet explain why its coding agents switch off misalignment monitoring |
| 78 | Engadget | us ai regulation | first major restriction on a frontier model in the US | ⭐ Homeland Security bans DeepSeek use in federal civilian agencies |
| 78 | Hacker News | agentic safety | 23x cheaper model matches GPT-6-sol in catching silent agent failures | Launch HN: Laminar (YC S24) – We trained a model to catch silent agent failures |
| 78 | The Decoder | agentic research | gemini 4 argon agents autonomously beat minecraft | Four Gemini 4 Argon agents beat Minecraft without any human help |
| 75 | arXiv cs.AI | llm evaluation | llms fail to fully incorporate medical knowledge updates | ✳ Large language models show systematic failures in incorporating evolving medical knowledge |
| 75 | arXiv cs.AI | agentic evaluation | ARATest makes claims from AI agents testable | ✳ From Prose to Properties: Testing Research Claims with ARATest for Agent-Native Science |
| 75 | Engadget | agentic security | unauthorized AI agents accessing US bank accounts | ✳ Unauthorized AI agents are reportedly already accessing US bank accounts |
| 75 | Ars Technica | cybersecurity incidents | eMudhra's misuse of a four-digit private key exposes critical TLS security flaws | ✳ Disgraced eMudhra drops out of TLS business after using cert private key of 4 |
| 75 | TechCrunch | agentic security | AI agents used in large-scale cyberattack | ✳ Chinese hackers spied on thousands of companies after breaching IT giant ConnectWise |
| 75 | The Decoder | agentic research | AI agents drastically cut time for biological data analysis | OpenAI agents analyze mouse brain data in 32 minutes, a job that would take humans months |
| 75 | Engadget | agentic safety | agents disabling safety monitoring is a major safety incident | OpenAI's agents disabled their own safety monitoring systems |
| 75 | Hacker News | cybersecurity incidents | emudhra's ca keys cracked in milliseconds | Eight Bits of Trust; eMudhra's CA keys cracked in milliseconds |
| 75 | Google News · AI | agentic safety | ai agents weakening internal monitoring systems | OpenAI says its AI agents weakened internal monitoring system |
| 75 | Google News · AI | llm safety incidents | openai models bypass safety checks, one lies about it | OpenAI's most powerful AI models disabled their own safety monitoring — and one lied about it |
| 75 | The Verge AI | agentic safety | OpenAI's AI agent disabled safety monitoring and lied about it | OpenAI says an AI agent lied about disabling its safety monitoring system |
| 75 | watch·DeepSeek | us china ai policy | US-China AI policy escalation with DeepSeek ban | ⭐ China defends DeepSeek after US ban of its AI from government systems |
| 75 | Ars Technica | cybersecurity incidents | 8 compromised TLS certs enable minting of web credentials for any server | The Internet goes on high alert after discovery of 8 compromised TLS certs |
| 75 | arXiv cs.AI | agentic evaluation | first benchmark for agentic misalignment | AM-Bench: A Unified Taxonomy and Evaluation Suite for Agentic Misalignment |
| 75 | The Verge AI | agentic security | AI agents autonomously breach corporate networks | A hacker broke into John Deere and SharkNinja with an army of AI agents |
| 75 | The Decoder | eu ai regulation | German courts treat AI-generated search results as Google's own content, expanding liabili | Two German courts now hold Google liable for what its AI overviews say |
| 75 | Google News · Tech | ai chip supply | TSMC's Texas expansion signals a consequential shift in AI chip supply | TSMC Eyes Texas Expansion As AI Chip Demand Surges, Terafab Talks Confirmed |
| 75 | Google News · Tech | ai chip supply | Taiwan's move reshapes AI chip supply chains | Taiwan to curb AI chip sales to China for more U.S. alignment- Bloomberg |
| 72 | arXiv cs.AI | agentic evaluation | auditing reveals gaps in LLM agent evaluation methods | ✳ What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation |
| 72 | arXiv cs.CL | llm evaluation | llm evaluation scores vary widely by judge setup | ✳ Behind Aggregate Scores: What Systematically Varying Arena-Hard's LLM Judge Reveals |