Microsoft co-founder Bill Gates says the world should brace for a wave of AI-driven cyberattacks over the next year. Speaking to CNN’s Jake Tapper on State of the Union on October 11, 2026, Gates warned that criminals are gaining access to powerful AI models without adequate safeguards, and the combination could produce what he called “a lot of negative surprises,” mostly in the cyber area. In this article we break down exactly what Gates said, the evidence backing his warning, and what it means for businesses and ordinary internet users.
What Gates said on CNN
During the interview, Tapper asked Gates about the risks of advanced AI. Gates pointed to three ingredients coming together at once: bad actors, no safeguards, and the latest models. His argument was blunt: when those three combine, the most likely fallout over the next twelve months is a surge of negative surprises in cybersecurity.
Gates said criminals could exploit the newest AI technology to disrupt financial systems, extort businesses, and carry out attacks that are more sophisticated than anything seen before. He explained why cyberattacks are the likeliest place for trouble to appear first: there is already a known business model behind them. Attackers threaten to cause harm and demand payment, or they create a disruption that moves markets or affects betting outcomes.
The bioterror question
Tapper pressed Gates further on a darker scenario: whether AI could make it easier for terrorists to develop deadly biological weapons. Gates has previously described the bioterror risk as potentially fifty times greater than that of a natural pandemic, and he did not back down on Sunday.
Asked whether world leaders are doing enough to prevent such an attack, Gates answered plainly: no, they are not doing enough. He compared unregulated AI with other powerful technologies society has learned to govern. Cars, planes, and drugs all carry real dangers, he said, but we built significant regulatory capability around them. Artificial intelligence, in his view, is more dangerous than any of those, and the guardrails are not keeping up.
Gates pushes back on the doomerism label
Tapper asked whether Gates’s concerns would fall under what OpenAI CEO Sam Altman has called doomerism. Gates declined the framing, saying that all responsible AI leaders, including Altman and Anthropic CEO Dario Amodei, are telling the world that AI is very concerning. The difference, in his account, is that alarmism about imminent human extinction goes too far, while specific near-term dangers like AI-powered cyberattacks do not get enough attention.
Amodei himself wrote a public letter on September 12 calling for a slowdown of AI development so safety and oversight efforts could catch up with the rapidly evolving technology. That call came just weeks before the White House brokered a voluntary AI safety pact with major tech companies, followed by a tougher AI incident reporting mandate in early October.
What the data says: AI-enabled attacks are already rising
Gates’s warning did not arrive in a vacuum. CrowdStrike’s 2026 Global Threat Report found that attacks by AI-enabled adversaries increased 89 percent year over year in 2025, while the average time criminals needed to move laterally through a compromised network fell to just 29 minutes. In other words, the surge Gates predicts is arguably already underway.
Recent cases show how real this has become:
- Investigators linked a Chinese-speaking hacker to cyberattacks on South Korean banks, using an AI security tool called ARTEX to automate parts of the campaign. The tool’s developer later withdrew its public source code after the reports surfaced.
- Japan declared a cybersecurity emergency after AI-powered attacks hit more than 20 major firms, including retail and automotive companies.
- Researchers have documented AI agents chaining together real attacks in the wild, from exploiting vulnerabilities in security software itself to filing a fabricated crime report with police through a website tip form.
These incidents suggest the line between experimental AI misuse and routine criminal tradecraft is already blurring.
What this means for businesses
For companies, the practical takeaway is that the threat model is changing. Phishing emails are becoming harder to spot because AI can write them flawlessly. Vulnerability discovery is accelerating as AI agents scan code faster than human researchers. And the cost of running a sophisticated attack campaign is falling, which puts smaller businesses in the crosshairs, not just large enterprises.
New industry efforts to use AI for defense, such as free vulnerability scanning programs for open source projects, show that defenders are getting the same tools. The race is now about speed: who finds and fixes the weakness first.
What ordinary users can do
You cannot fix the lack of global AI safeguards yourself, but you can reduce your own exposure to AI-powered attacks:
- Turn on multi-factor authentication for email, banking, and cloud accounts. AI-written phishing is convincing, so a second factor is your strongest defense.
- Keep devices and software updated. Many AI-accelerated attacks exploit known, unpatched flaws, and the patch guidance still applies regardless of how the attack was built.
- Treat urgent money requests, even from familiar voices or writing styles, with skepticism. AI can clone both.
- For business owners, segment your network and keep offline backups so one compromised machine cannot take down everything.
FAQ
What exactly did Bill Gates warn about on CNN?
On the October 11, 2026 episode of State of the Union, Gates said the combination of bad actors, missing safeguards, and the latest AI models is likely to produce many negative surprises over the next year, primarily in cybersecurity. He also reiterated that AI could make bioterrorism far easier and said world leaders are not doing enough.
Why does Gates think cyberattacks will come first?
He argued that cybercrime already has a proven business model: threaten disruption and demand payment, or cause disruption that moves markets. AI simply makes that model cheaper and harder to stop.
What does the data show about AI-powered attacks?
CrowdStrike’s 2026 Global Threat Report found an 89 percent year-over-year rise in attacks by AI-enabled adversaries in 2025, with the average lateral-movement time down to 29 minutes. Reported cases include AI-assisted attacks on South Korean banks and a string of AI-powered intrusions at more than 20 Japanese companies.
Does Gates support slowing down AI development?
He did not call for a freeze in this interview, but he aligned himself with other AI leaders who say safety concerns are serious. Anthropic CEO Dario Amodei went further on September 12 with a public letter urging development to slow so oversight can catch up.
How is AI connected to the White House safety pact?
Weeks before Gates’s warning, the White House brokered a voluntary AI safety accord with major tech firms, then followed it with a mandate requiring immediate disclosure of AI incidents. Gates’s comments add weight to the argument that voluntary measures may not be enough.