Open-weight AI models are catching up to the frontier, analysis finds

The leading open-weight AI models are closing the gap on the frontier in cyber capabilities, according to the UK’s AI Safety Institute.