# AI: Neither Tortoise Nor Hare!

# AI: Neither Tortoise Nor Hare!

Out-of-control artificial intelligence (AI) agents are making headlines. The widely publicized attack on Hugging Face by OpenAI agents has been extensively discussed and analyzed in highly anthropomorphic, alarmist terms. There has been talk of a “gang” of agents, “ringleaders,” and even agents “sacrificing” themselves for their shared cause.

So, is Frankenstein lurking in our IT systems? We must admit that we can’t really explain when or why some agents disregard the guardrails put in place. Fortunately, most companies have not deployed autonomous agents, but rather agents restricted to predetermined sequences of actions. The risk of these agents “taking control” is therefore very limited.

What, then, of the AI race? Should we slow down and risk becoming the tortoise, or keep moving at the hare’s pace?

In the short term, there is no question: every company needs to sprint on cybersecurity. Attack tools have become extremely powerful, and the defenses are not yet in place. Data breaches across every continent are becoming more frequent and more public. Companies selected for early testing of Mythos are discovering numerous vulnerabilities in record time. Here, accelerating is the only option.

Beyond that, enterprises will keep up the pace of AI deployment, though this is more of a middle-distance race than a sprint. Competition demands an intense pace, but the goals are medium-term. Forty-two percent say they intend to deploy more autonomous AI systems by 2030, so they have to learn how to control them. As one AI researcher told me, “An agent is a bit like a lion: you don’t know exactly what’s going on inside its head. Even if it has been trained, meaning guardrails have been put in place, you’re better off keeping it in a cage if you have it in your home.”

Finally, when it comes to research, the decision to slow down or speed up rests with the labs and their teams. A moratorium is hard to envision.