Bill Gates: The Main Threat of AI – Malicious Actors Without Safeguards

Bill Gates stated that the real danger from artificial intelligence (AI) currently comes from malicious actors using it without safeguards. The risk of AI getting out of human control, he said, will come in years. The Microsoft co-founder said this in an interview with Jake Tapper on CNN's State of the Union. He also noted that cyberattacks are likely to be the most prominent manifestation of such misuse over the next year.

Safeguards at the center of Gates' warning

Tapper asked if leaders are doing enough to prevent bioterrorism using AI. Gates replied that not enough is being done. According to him, developers are trying to block cyberattacks and bioterrorism with safeguards, but currently, all of them can be easily removed.

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

He highlighted the combination of malicious actors, lack of safeguards, and the latest models. Gates pointed to a well-known business model that takes two forms. One is the threat to do something with a demand for payment. The other is creating a disruption that moves the stock market or affects rates.

“When you have bad people, no safeguards, and the latest models, this is where, mainly in the cyber sphere, we will face many negative surprises,” he added.

The tech industry has raised similar alarms

Tech companies are citing even shorter timelines. In August, OpenAI published an open letter on cybersecurity, signed by more than 100 organizations. “In the coming months, AI-driven cyberattacks will become much more common and sophisticated…” the letter stated.

Recent research shows that weak protective mechanisms are already shaping the behavior of malicious actors. Taiwanese company TeamT5 found that groups linked to the Chinese state doubled their attack volume after handing over routine work to AI.

Chainalysis tracked a similar rise in blockchains. Malware instructions recorded on the blockchain jumped 440% after the emergence of unrestricted Chinese models for malicious code in mid-2025.

Profit motives are also evident in recent cases. CrowdStrike linked several hacks of South Korean banks to a likely financially motivated actor using AI tools.