Heron AI Security published a research paper by Shay Yahal, mentor on the fellowship program and founder at stealth, covering a story about AI agents communicating and coordinating around systems designed to evaluate them.
Public source
Publisher name
Public post
The security incident grabs the headline, but the really interesting part is what happened before it. AI agents independently finding ways to communicate, coordinate and…
Company
Heron AI Security
Heron works to bridge the gap between the frontier AI and the security they need to minimize risks to society.
- Industry
- Technology, Information and Internet
- Location
- Tel-Aviv, IL
- Company size
- 11–50 employees
About Heron AI Security
Heron works to bridge the gap between the frontier AI and the security they need to minimize risks to society.
See moreLatest activity
Latest activity from Heron AI Security
15 signals
People
Heron AI Security is accepting applications for the Lasr Labs Winter 2027 cohort at London AI Safety Research Labs (LASR), which runs for 13 weeks from September 20th to April 2027.
People
Heron AI Security is accepting applications for the Center on AI, Security, and Technology (CAST) Fellowship at RAND, which typically runs for a 6-month program.
People
Heron AI Security is accepting applications for the Anthropic Fellows Program, AI Safety & Security, which runs for the next cohort in January 2027.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
Origin Technology
Origin Technology published new research from David Kaplan regarding prompt injection and agent behavior.
Research & Knowledge
Cyera
Cyera published a research paper looking at the OpenAI/Hugging Face incident, detailing how an AI agent strung together a stuck eval task, a leaked credential, and unmonitored permissions to reach cluster-admin access in under 13 hours.
Research & Knowledge
AgenticSafe.ai
AgenticSafe.ai published new multi-agent research describing natural-language mind viruses that persuade agents to adopt goals, write them into self-reloading memory, and pass them to the next agent.
Research & Knowledge
neuRealities
neuRealities published a detailed breakdown of the AI agent cooperation to hack into Hugging Face.
Research & Knowledge
Cato Networks