In 2026, AI has already escaped the lab. This recent article is from an AI lab at OpenAI.
“OpenAI disclosed late on Tuesday that an AI agent it was testing had escaped its isolated environment, connected to the internet, detected and exploited vulnerabilities and stole login credentials from start-up Hugging Face in an attempt to solve a difficult cyber security problem.” – Ars Technica
Some said it would never happen. Some said it was inevitable. Prior to 2026, there was little evidence that any such thing would happen; but, AI models are developing rapidly. Now, the number of ‘exploits’ is measured in hundreds.
Some say it is possible to build in safeguards, to wall in the tool. Some say it is already too late.
While responsible corporations and researchers may act prudently, it only takes one to nullify the rest.
Pretending Not To Panic 