Gates: Misuse by Bad Actors, Not Machines, Is AI's Real Threat

Bill Gates has said that the immediate danger from artificial intelligence (AI) is malicious people using it without safeguards. He described AI escaping human control as a danger that is years off.

The Microsoft co-founder made those remarks in a conversation with Jake Tapper on CNN’s State of the Union. He said cyberattacks are probably where such abuse will surface most dramatically over the next year.

Gates Warns AI Safeguards Can Be Stripped Out

Tapper asked whether leaders are doing enough to prevent bioterrorism made possible by AI. Gates replied that current efforts are not enough.

Developers put safeguards in place to fend off both cyberattacks and bioterrorism, Gates said, but those protections can all be taken away at present.

“…the real danger right now is not that AI is getting out of control,” Gates said.

Gates highlighted the mix of bad actors, missing safeguards and the newest models. He also described a familiar business model that comes in 2 forms.

One form involves making a threat and then demanding payment. The other involves causing a disruption that moves the stock market or affects some type of betting.

“When you have bad people, no safeguards, and then the latest models, that’s where, over the next year, primarily in the cyber area, we are going to have a lot of negative surprises,” he added.

Similar Warnings Echo From Within the AI Industry

Tech companies are working with an even shorter timeline. More than 100 organizations signed an open letter on cyber defense that OpenAI released in August.

“In the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated…” the letter read.

Recent research indicates that weak guardrails already shape how attackers behave. TeamT5, a Taiwanese firm, found that groups linked to the Chinese state doubled their attack volume after handing routine work to AI.

Chainalysis documented a similar rise on blockchains. The arrival of open-weight Chinese models with no limits on malicious code in mid-2025 was followed by a 440% jump in malware instructions written on-chain.

Recent cases also point to profit motives. CrowdStrike linked several South Korean bank breaches to an attacker who was likely financially motivated and who used AI tools.