Perspectives

Positions we are willing to defend, on where AI trust, security and regulation are heading — and what that asks of the people building with them.

Perspectives14 October 2026
Terry Vogelsang
Anthropic's Self-Replicating AI Research: What Security Teams Should Do Now
What Anthropic's research on self-replicating AI means for enterprise security, autonomous agents, and continuous AI assurance.
Read it
Perspectives24 July 2026
Adrien O'Hana and Prof. Andrei Kucharavy
Nobody should audit their own containment: lessons from the OpenAI and Hugging Face incident
The OpenAI–Hugging Face incident shows how weak containment, absent monitoring and poorly balanced safeguards turned predictable agent behaviour into a security breach.
Read it
Perspectives9 June 2026
Adrien O'Hana
Mythos, nine weeks on
Mythos is not much better at offensive cyber than the models beside it. What changed is the entry price — and the answer is ordinary defensive discipline, run at the speed the threat now moves.
Read it
Perspectives15 April 2026
Gaetan Stein
Surelio is listed in the OWASP GenAI Red Teaming Landscape
The OWASP GenAI Security Project has published its AI Security Solutions Landscape for Q2 2026, and Surelio is listed across six of its nine lifecycle categories.
Read it
Perspectives2 September 2025
Gaetan Stein
Independence in AI safety audits: a requirement, not a courtesy
AI safety audits require the same independence as financial audits — a consulting firm cannot be both judge and jury on a system it helped build.
Read it

Let's talk

Let's engineer confidence together.

Tell us what you're building, securing, or trying to solve. We'll help identify the right next step.

A 30-minute call, no slideware
We tell you which engagement fits — or that you don't need one yet

Prefer a conversation? Book a 30-minute call