Following news of the White Houses’ discussion with leading AI companies to review a newly completed voluntary framework framework for testing the cybersecurity capabilities of advanced AI models, Mike Britton, CIO at Abnormal AI, had the following to say :

“I’ll admit to some scepticism about who benefits most from this kind of framework. The companies with the scale to absorb a compliance process are often the same ones who’d like to see that same process become a barrier to anyone smaller catching up. That doesn’t mean the underlying concern is wrong.

There does need to be real risk analysis on what a new model is capable of, good and bad, before it ships — and it should include serious investment in red-teaming and pressure-testing, not just a paperwork exercise. The question is whether this framework is designed to produce that rigor or just to look like it.“