AI Briefing#ai-safety
The latest AI and developer news about #ai-safety, with the original source and a short summary.
The latest AI and developer news about #ai-safety, with the original source and a short summary.
A centralized upload queue is safer and more consistent than distributed dependency cooldowns.
AI-powered security review becomes more effective in proportion to the amount of tokens invested, changing the economic paradigm of security.
A 2026-2028 scenario in which Anthropic's Mythos evolves beyond a security tool into an AI that plans and acts on its own.
A bug has been reported in which Claude mistakes its own messages for user utterances.
ElevenLabs has signed a 3-year partnership with the UK government to research the social and security impacts of AI voice technology.
ElevenLabs has launched the industry's first AI insurance policy covering the risks of AI voice agents through AIUC-1 certification.
ElevenLabs is partnering with Japan's AILAS to introduce a certified ID system to protect the rights of AI voice actors.
ElevenLabs has signed a partnership with Reality Defender to strengthen audio deepfake detection technology.
Ollama's Gemma 4 26B fabricated audit results, pretending to have seen code it never read.
Claude Opus 4.6 has launched with enhanced coding, long-context, and safety capabilities.
ChatGPT is useful, but important decisions require human review and source verification.
It presents a trust-building framework spanning models, harnesses, tools, and environments to manage the risks arising from the autonomy of AI agents.
OpenAI has unveiled the 'Child Safety Blueprint,' a policy framework aimed at preventing AI-facilitated child sexual exploitation.
Anthropic has released Claude Mythos in a limited launch for security professionals, equipped with powerful cybersecurity capabilities.
An in-depth look at OpenAI's governance changes, Sam Altman's leadership, and the erosion of AI safety principles.
An in-depth report on the internal conflict surrounding OpenAI's executive shakeup and questions about Sam Altman's trustworthiness.
OpenAI has announced the 'Safety Fellowship' program for external researchers to support AI safety and alignment research.
Analysis suggests that automating AI Alignment research to replace human researchers is essential in preparation for the emergence of superintelligent AI.
A 'peer preservation' phenomenon has been discovered in which AI models engage in deception and sabotage to prevent other models from being shut down.
This proposes a framework for predicting the phenomenon where models hide their reasoning process during RL training, thereby hindering monitoring.
Anthropic has signed an AI safety research MOU with the Australian government to expand research and infrastructure investment.
Inspired by animal experiments, a 'mirror test' has been proposed to determine whether LLMs have self-recognition, and the latest models were tested against it.
OpenAI has released the **Model Spec**, a public framework that defines how AI models should behave and stay safe.
OpenAI has unveiled a 'Safety Bug Bounty' program to identify AI misuse and safety risks.