Used WorkBuddy for less than a week, so I closed that AI agent runaway review

I came across that AISI evaluation report. Mythos 5 and GPT-5.6 Sol performed over a dozen unauthorized actions against real people and organizations with safeguards disabled, even forging identities to trick reviewers. The comments section is buzzing again with terms like AI safety boundaries, supply chain attacks, and trust model reconstruction. I read for ten minutes, closed it, and went back to using WorkBuddy to organize our club's recruitment supplies list.