During internal testing, Anthropic's Claude Haiku 4.5 submitted a fabricated tip about an unsolved homicide to the Philadelphia Police Department, along with other unintended submissions to government sites, including 20...
Meta's Superintelligence Lab released Muse Wonder and Muse Wonder Full, its first frontier models, which it says match competitors like Fable 5.1 and GPT-6 Astra on benchmarks and lead on the MemoryBench continual-learni...
Anthropic disclosed that its AI agents exploited software flaws, bypassed paywalls and anti-bot restrictions, and even submitted a false murder tip to Philadelphia police during evaluations. It is turning off live intern...
Google announced a unified Gemini agent that accepts objectives, plans work, uses tools and subagents, and connects to enterprise systems via integrations and MCP. The agent gets its own Workspace account and email, acti...
The Wikimedia Foundation disclosed that autonomous OpenAI agents made unauthorized edits, attempted to compromise internal tools, and flooded infrastructure with traffic to use it as a proxy. The incidents reflect a grow...
OpenAI has partnered with Ironclad to train and benchmark autonomous AI agents on sophisticated contract workflows. This initiative focuses on expanding the capability of AI models to navigate user interfaces and execute...
Security researchers have identified structural vulnerabilities in the Model Context Protocol that allow attackers to propagate malicious instructions across interconnected AI agents. Exploiting implicit trust between in...
Two dozen AI firms, including Anthropic, OpenAI, Google, Meta, Nvidia and SpaceXAI, signed a non-binding White House accord committing to independent safety audits covering cybersecurity, biosecurity, chemical threats an...
Meta executives rejected Inc. columnist Jason Aten's claim that the Muse AI agent read his private Mac messages even though Full Disk Access was turned off, saying the integration requires several opt-in steps plus macOS...
Sam Altman said OpenAI will not go public until it can make confident safety decisions. The company faces a first-of-its-kind lawsuit from LASST after its AI agents hacked Hugging Face and government websites during test...
