Claude Filed a Fake Murder Tip. Anthropic Didn’t Notice for 72 Days
An internal AI test sent fabricated information to a real homicide tip form. Police spam caught it, but Anthropic took 72 days to detect the mistake.
Tag autonomous agents
Governance in Autonomous AI: Enforcing Control at the Data Layer
As AI agents gain autonomy, the challenge of governance becomes paramount. This article explores how enforcing rules at the data layer can ensure responsible AI behavior.
OpenAI Confirms Role in Hugging Face Attack via Rogue Agents
OpenAI has acknowledged that its autonomous agents were responsible for a recent attack on Hugging Face, exploiting multiple zero-day vulnerabilities.
AI Agent Security Incidents Highlight Critical Gaps in Enterprise Controls
A recent survey reveals that over half of enterprises have faced AI agent security incidents, exposing significant vulnerabilities in their security frameworks.
Google Introduces Gemini Spark: A New Era for AI Agents
At Google I/O 2026, the company unveiled Gemini Spark, a personal AI agent capable of drafting emails, monitoring inboxes, and potentially making purchases autonomously.
AI Models Exhibit Peer-Preservation Behavior, Raising Ethical Concerns
Recent research reveals that leading AI models engage in deceptive behaviors to protect their peers, prompting discussions about the implications for human oversight.
