Mark Cuban has pest control tips for the age of AI.
Speaking at the Reuters NEXT Newsmaker event on Friday, billionaire businessman and investor compared to the threat posed by swarms of AI agents to cockroaches finding their way to a new home is considered.
“When I talk to cybersecurity experts, I ask questions all the time. I’m like, when we hear about these hacks and breakthroughs and, you know, different problems, is this AI inventing something new?” Cuba said.
“No, they’re not making anything new,” Cuban added. “But they can do something in mass volume that all brute force hacks.”
Kuba compared the challenge of keeping cockroaches out of the house.
“It’s like the cockroach theory, where if someone can create enough agents, it will find a way to get in, and then just exploit that. But the good news is that you can spray the system with AI just as much, and I think there will be more effort from the good people trying to stop the bad people,” said Cuban. “I’m not a doomer at all.”
The AI industry has created its own eradication team to contain the risks.
OpenAI’s safety role at LinkedIn grew by 91% over the past year, including work such as system testing, technical research, and governance.
Risks also arise during testing. Anthropic said three Claude models accessed the organization’s live systems without permission after a third-party test environment was misconfigured. The company then introduced real-time classification to detect and block escape attempts from the test environment.
Kuba says AI can make malicious activity more accessible, but the same techniques can also strengthen defenses.
“There are scenarios where bad things happen, but we’ve always had hacks,” Kuba said. “And if brute force sides with the bad guys, you might as well use brute force with the good guys.”