**Regulatory/safety landscape: cyber fears, EU Act, bans, AI errors, war crimes** [developing]
Key Questions
What is Anthropic’s Project Glasswing?
Project Glasswing is an AI model designed to fight AI-powered cyberattacks. It gates access to defensive capabilities first. It addresses rising cyber fears in the regulatory landscape.
How has Japan changed its privacy laws for AI?
Japan relaxed privacy laws to become the 'easiest country to develop AI' by allowing more data use. This supports AI training without strict prior restrictions. It contrasts with global safety concerns.
What fears surround Manus AI agent?
Manus AI agent raises fears of a 'deepfake' explosion online through agent swarms. Users need awareness of its demo capabilities. It ties into broader deepfake and privacy risks.
What safety issues were found in Kimi K2.5?
A new paper reveals concerning dual-use capabilities in Kimi K2.5, questioning its safety and alignment. It highlights risks like those in OpenClaw. This fuels regulatory scrutiny.
What is AgentHazard benchmark?
AgentHazard evaluates harmful behavior in computer-use agents. It joins AgentSocialBench for privacy risks in agentic networks. These address AI errors and potential war crimes concerns.
What actions did Governor Newsom take on AI?
Governor Newsom signed an executive order to strengthen AI protections and responsible use in California procurement. It responds to hallucinations, suits, and reg needs. This shapes the US regulatory landscape.
What happened with Meta's AI data work?
Meta froze AI data work after a Mercor breach risked training secrets. Mercor, a $10B startup, faced a major security incident. This underscores data poisoning and ModelSpy-like risks.
What is ModelSpy and related attacks?
ModelSpy refers to attacks stealing models without system access, like new AI theft methods. It ties into US AI war targeting and EU Act developments. These amplify cyber and safety crises.
Anthropic Glasswing cyber defense gating; Japan privacy laws relaxed for AI data; Manus deepfake agent swarms; prior Kimi/OpenClaw risks, US AI war targeting, suits, Meta poison, ModelSpy, reg (EU/Newsom/etc.), hallucinations.