MIT Tech Review AI highlights Google DeepMind's $10 million initiative to study multi-agent AI systems, but the effort may be insufficient given the scale of the potential risks. While the company aims to preempt scenarios like AI-driven scams or cyberattacks, critics argue that the funding is modest compared to the complexity of the problem. Moreover, the reliance on academic research, while valuable, may not yield actionable solutions fast enough to keep pace with industry deployment. The real test will be whether these findings can translate into practical safeguards before AI agents become ubiquitous.