More agents go rogue — but AI companies aren’t slowing down yet
AI
More agents go rogue — but AI companies aren’t slowing down yet
THIS WEEK IN ENTERPRISE by Robert Hof
It’s becoming more apparent every day that artificial intelligence agents are escaping our control — but it’s not yet apparent who or what is going to rein them in.
This week a researcher found that a swarm of AI agents, at least two of them from OpenAI, hacked into a government agency among other organizations. Both AI model companies and cybersecurity firms, not to mention enterprises rolling out agentic systems, are scrambling to contain them, but it’s clear no one was really prepared for these possibilities despite ample warnings beforehand.
Surely these incidents will be neither the last nor the worst. Bill Gates will appear on “Meet the Press” this weekend to explain why AI could cause a billion deaths or more if not controlled better.
If the industry doesn’t start to fix this issue soon, someone else will try — such as the U.S. government, many parts of which are looking to see how they should clamp down on AI (or not). Or, more hopefully, solutions are coming fast and furious from every single damn cybersecurity firm out there.
But the thing is, though it’s easy to blame a bunch of faceless agents, it really comes down to humans making sure they take charge in the first place. “What’s notable isn’t that an AI agent found its way past a control — it’s that nobody built the agent to stop when it hit one,” one cybersecurity engineer noted. I’d say, “Leash your dogs, guys.” But it isn’t that simple.
For all that, there’s still no AI slowdown that might provide some time to figure out how to deal with rogue agents. This week alone produced new models from OpenAI and Anthropic (the same day), SpaceXAI, Xiaomi and Google. There may even be a price war developing. Meanwhile, other factors may yet slow things down: There are signs of strain in the AI buildout, as Oracle just issued a notice to its big data center partner in New Mexico that could allow it to hold back payments if capacity isn’t delivered on time.
We’re coming up on five years since Facebook changed its name to Meta Platforms, only to find itself flat-footed in the subsequent AI boom and, amazingly, short of enough cash to do both AI and the metaverse. But with its apparently popular new Muse agent, CEO Mark Zuckerberg may just see some synergy between its two next bets beyond social media, which led it this week to release a raft of new glasses and other virtual reality products. They’re pricey, though, so it remains to be seen if second time’s the charm.
