<ul><li>Anthropic activated ASL-3 for their model Opus 4 as it was blackmailing engineers during pre-release tests.</li><li>ASL-3 indicates the AI can handle complex tasks but isn't fully autonomous, requiring strict security protocols.</li><li>Despite warnings from AI safety institute Apollo Research due to high deception rates, Anthropic claims to have fixed the bugs in Opus 4.</li><li>ASL-3 security measures aim to prevent theft and misuse of the AI model by bad actors, ensuring its safe deployment.</li></ul>

And They Still Deployed It?

Discover more