‘I Hate Being the Bearer of Concerning News’ — Bill Gates Warns Even the Best Case for AI Means a 20-Year Transition With Winners and Losers

Artificial intelligence has a safety problem that can’t be solved by a company’s terms of service. Microsoft Corp. (NASDAQ:MSFT) co-founder Bill Gates says the technology has advanced to the point where government oversight is no longer optional, warning that powerful AI models could be exploited to cause serious harm.

"I hate being the bearer of concerning news on this, but it’s even worse not to tell people that this is coming and we have choices as humans to make about how we shape it," Gates said on the "Decoder" podcast last week.

The warning marks a more urgent tone from Gates, who once emphasized AI’s potential to improve lives while urging developers to guide its progress responsibly.

"Overall, I believe that it’s a beneficial thing, and we need to just shape it in the right way," Gates told The Verge in 2024.

Gates Says AI Safety Can’t Be Left To Tech Companies

In the latest interview, Gates told The Verge editor-in-chief Nilay Patel that the problem isn’t simply whether one AI model is more dangerous than another. It’s whether powerful systems can be stripped of safeguards and used to carry out attacks.

"The risk isn’t one model versus another model," Gates said.

He warned that models without adequate protections could help cyberattackers or bioterrorists cause enormous damage. AI systems capable of finding software vulnerabilities or reasoning through complex biological problems could be useful to researchers, but those same capabilities could be abused.

Gates said the issue requires safeguards that go beyond voluntary promises from the companies developing the technology.

"We are faced with a huge safety problem," Gates said, arguing that governments must work with industry experts while ensuring that models retain protections against misuse.

The Threat Is More Immediate Than Science Fiction

The debate over artificial general intelligence often centers on whether machines could eventually become smarter than humans. Gates argued that society doesn’t need to settle that question before addressing the risks.

The immediate concern is whether AI can be used to attack critical systems, including banks and electricity grids.

"There is a business model for cyberattacks," Gates said.

In other words, malicious actors don’t need to build a conscious machine or achieve some futuristic milestone to cause damage. They need tools that make attacks easier, faster or more effective.

Gates also pointed to the danger of AI systems whose safeguards can be removed. That creates a challenge for policymakers, who must consider not just what a model can do but also how reliably its protections hold up when someone tries to bypass them.

Government Faces Pressure To Set The Rules

Gates argued that competing AI companies cannot be expected to manage every risk on their own. Governments need to establish baseline safeguards while drawing on technical expertise from the industry.

That approach raises difficult questions about how to regulate powerful models without blocking useful research or giving malicious actors access to dangerous capabilities.

The goal, Gates suggested, should be to ensure that AI’s benefits continue to develop while reducing the chance that its most powerful tools become weapons.

Gates remains optimistic about what AI can accomplish in medicine, education and other fields. But optimism alone won’t protect the public. As the technology grows more capable, the challenge is making safety measures keep pace.