Research — Verno Labs

Heading to Black Hat?Read more

Heading to Black Hat? See offensive AI security in action.Read more

Platform Research Docs

ResearchAI agents are being deployed faster than we know how to test them. Verno Labs Research exists to close that gap. We focus on frontier red teaming for agentic applications: discovering novel attack classes, building evaluation methodologies that scale, and publishing the benchmarks and tools the ecosystem needs to ship safely.

ResearchAI agents are being deployed faster than we know how to test them. Verno Labs Research exists to close that gap. We focus on frontier red teaming for agentic applications: discovering novel attack classes, building evaluation methodologies that scale, and publishing the benchmarks and tools the ecosystem needs to ship safely.

JUL 15, 2026Stella Katsarou, Member of Technical Staff

VernoBench: Rethinking prompt extraction evaluation

"What sets Verno Labs apart is their deep production experience combined with a cutting-edge research DNA. This positions them perfectly towards delivering a future-proof solution to agentic security."

Marios Kogias, Assistant Professor

JUL 15, 2026Stella Katsarou, Member of Technical Staff

VernoBench: Rethinking prompt extraction evaluation

"What sets Verno Labs apart is their deep production experience combined with a cutting-edge research DNA. This positions them perfectly towards delivering a future-proof solution to agentic security."

Marios Kogias, Assistant Professor

Articles

JUL 9, 2026Verno Labs and Imperial College CTF MAY 27, 2026What the Composio incident actually shows MAY 22, 2026Guardrails caught the known attacks. They missed the new ones.MAY 6, 2026Anthropic builds the best hacker model, then locks it away APR 14, 2026Your AI agents think out loud: that's a security opportunity — for now.MAR 12, 2026The morning routine that hands over your accounts

Articles

Verno Labs and Imperial College CTF

What the Composio incident actually shows

Guardrails caught the known attacks. They missed the new ones.

Anthropic builds the best hacker model, then locks it away

Your AI agents think out loud: that's a security opportunity — for now.

The morning routine that hands over your accounts