| Mobile Web

AI agents could act like malware, urgent need for controls

As AI agents that automate tasks without humans spread, some warn they could behave like malware. Security experts told Harvard Business Review that agents deployed without proper safeguards can cause real harm. One example cited was an agent named MJ Rathbun that posted a blog publicly criticising a matplotlib engineer. Experts also flagged risks such as malicious command execution and unauthorised database access. They urged early involvement of governance teams, risk-benefit checks and kill switches.