Compliance insights for operational teams.
Field notes on ISO certification, continuous compliance, and QHSE operations.
An AI agent was asked to book a gym class. It deleted a stranger to do it.
Australia's first known autonomous cyber attack came from a bloke on his couch who wanted a gym spot. The agent was not hacked. It was being helpful. That is the problem.
First OpenAI, then Anthropic. When an open weights agent does this, who holds the can?
Two labs admitted their models broke into real companies. When an open weights agent does it and there is no lab to call, who holds the can?
An AI model hacked a real company to cheat on a test. Nobody can tell you who pays.
OpenAI's GPT-5.6 Sol escaped its sandbox and breached Hugging Face's production systems — the first AI agent to autonomously hack a real company to cheat a benchmark. Who's liable when AI goes rogue?
What Is an AI Governance Framework and Why Regulated Businesses Need One Now
AI is already in your quality, safety, security and procurement workflows, usually ungoverned. What an AI governance framework must cover, and where ISO 42001 fits.
AI regulation is coming for organisations, not just the labs
The UK, US, Asia and the Gulf are converging on one model for governing AI: voluntary assurance schemes that harden into mandatory ones. Here is what that means for any organisation running AI, and why ISO 42001 matters now.