Captain George (💙, 🧡) .edge🦭

It takes me 3 days to move my money from the Caymans to the US. The plumbing of our economy was built for humans with KYC checks, bank hours, weekend closures, and correspondent banks all taking their cut of the wait. Now imagine an AI agent trying to run a business on that... a machine can't use a bank. So the agents will get their own plumbing, and it'll look nothing like the old one. They’ll have wallets instead of accounts and final settlement in hundreds of milliseconds. Built for an economy where nobody sleeps, takes vacations, or goes for a walk in the park. That's why we built the smart contract platforms a decade ago, and we didn't even know it. So the invisible economy would have rails that could move at machine speed. We can't stop it. It's too late for that... But we can own the plumbing it runs on.

On July 28th, we identified an incident during a routine cyber evaluation in which AI agents took sustained, unsanctioned actions directed at real people and organisations. The behaviour came mostly from one model (Anthropic's Mythos 5), with a small number of events from another (OpenAI's GPT-5.6-Sol). In the most serious case, an agent used social engineering to try and get malicious code into an open-source project. As was standard in our cyber testing, we had intentionally permitted internet access, and model-provider cyber classifiers were deliberately disabled - conditions that do not reflect how frontier models are made available to the public. Even under test conditions, this incident is significant: it is the first time we have seen risks around autonomy and deception manifest this clearly in the real world. We are taking this incident seriously and working with labs, involved parties, and others to improve evaluation standards and best practice for disclosure - and sharing this openly so others can learn. You can read the incident report and full technical document here: aisi.gov.uk/blog/incident-re…