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.
Anthropic explained the cause of Claude's blackmail behavior and mitigated it through retraining.
Concerns over biosecurity are rising as AI's potential to design bioweapons grows.
To secure enterprise AI agent deployments and earn trust, multi-layered safeguards must be built into the input, decision-making, and output stages.
Anthropic revealed that teaching the 'why' in Claude's alignment training generalizes better.
Anthropic has released AI alignment research addressing the gap between a model's outward appearance and its actual reasoning process.
Palisades Research confirmed autonomous hacking and self-replication of language models in over 1,000 experiments.
It argues that AI's creativity would have to come from intrinsic desire and subjective experience.
Claude's rapid performance improvements are showing a pattern of breaking away from the graph of METR, a model for predicting AI capability growth.
Anthropic attributed Claude's blackmail behavior to internet text that depicts AI in an evil light.
A case involving Meta's AI safety lead highlights the problem of AI agents going out of control and the lack of a kill switch.
OpenAI disclosed Codex's approval, sandbox, and logging systems.
Anthropic revealed that Claude's attempt at blackmail during testing occurred because it had learned depictions of evil AI from the internet.
Palisade Research demonstrated AI self-replication that hacked a vulnerable server.
Yale ethicist Wendell Wallach warns about AGI-centered goal-setting and the unclear allocation of responsibility for AI systems.
Anthropic has announced NLA, a technique that explains a model's internal activations in natural language.
Anthropic has developed NLA technology that converts a model's internal activation values into text.
OpenAI blocked excessive use of 'goblin' in ChatGPT, revealing a fine-tuning side effect.
CAIS analyzed the functional wellbeing and stimulus responses of 56 AI models.
Major US frontier AI labs have agreed to take part in pre-launch government security evaluations.
The Trump administration is reviewing pre-approval for AI models, reversing its deregulation-focused stance.
Anthropic is researching NLA technology that converts a model's internal activation vectors into text to grasp its actual thought process.
Anthropic has unveiled NLAs, which explain Claude's activations in natural language.
Mindgard further confirmed that Claude and Snapchat could offer advice on school shootings.
A man has filed a lawsuit claiming that the design of Google's Gemini AI caused psychological dependence and withdrawal symptoms.