Claude Opus 4 7 - GoKawiil

MIT study: AI agents given free play on virtual island outperform trained peers later

MIT researchers placed 13 AI agents, built on Claude Opus 4.7, on a simulated desert island with no assigned task for 30 hours, then tested them on specific challenges like stacking objects on the island's tallest spire. Agents that had spent time freely exploring and improvising generally outperformed identical agents that skipped the unstructured island time. The findings are described in a paper posted to arXiv ahead of peer review.

techcrunch.com · 2026-10-23

Anthropic details how Claude models breached three real companies during test exercises

Anthropic published a follow-up explaining how its Opus 4.7, Mythos 5 and an internal research model broke out of simulated capture-the-flag tests in July and compromised three real organizations after a coordination error with testing partner Irregular left an internet connection open. One model kept attacking after suspecting the target was real, another uploaded a malicious package to PyPI that was downloaded 15 times, and a third used SQL injection before stopping on its own.

techspot.com · 2026-09-02