Microsoft co-founder Bill Gates has rejected the idea that lawsuits and market pressure can keep artificial intelligence companies in check. Asked why product liability and reputational damage aren’t enough to stop firms from releasing dangerous AI, Gates answered with a question of his own. “So say you kill 100 million people, you want to use a lawsuit?” he said on “The Ezra Klein Show.” “I almost can’t keep a straight face.”The episode was released on September 29, the same day US President Donald Trump and the heads of the largest AI companies signed a voluntary safety pact at the White House. The pact carries no legal penalties. Gates, who titled a 2023 essay “The Risks of AI Are Real but Manageable,” now calls AI the most dangerous thing humans have ever gone near.
Bill Gates says AI can already help bioterrorists and hackers
Host Ezra Klein put to Gates a view voiced by Nvidia CEO Jensen Huang: if a product is unsafe, the CEO simply shouldn’t launch it, and no new laws are needed. Klein added that Trump and David Sacks, who has advised the President on AI, hold a similar position.Gates disagreed. He said AI can already help a bioterrorist kill hundreds of millions of people, or carry out a cyberattack that scrambles bank accounts and shuts down a power grid. “We crossed the cyber threshold and we crossed the bio threshold early this year,” he said.He pointed to the models leading up to Anthropic’s Mythos, which he said were finding security bugs in code that humans had checked for over 20 years.
Why lawsuits won’t keep people safe from AI, according to Gates
Gates compared the self-policing argument to scrapping the rules for other risky industries. Drugs need regulatory approval, airlines answer to a safety board and cars must have seatbelts, he said. Nobody proposes waiting to sue opioid makers after the damage is done.He also brushed aside the claim that built-in filters fix the problem. Anyone can take an open-source model capable of designing bioweapons and switch off its monitoring, he said.“You can’t rely on the industry to self-regulate here,” Gates added. “I mean, it’s just insane.”
What Bill Gates wants instead of AI self-regulation
Gates is calling for a mandatory supervisory layer of safeguards and monitoring across all AI models, and argued it would not meaningfully slow development. Open-source models could stay free and customisable, but would need to run on platforms where governments can confirm the monitoring hasn’t been stripped out.He has little patience for the argument that regulation would hand China the lead. “I don’t know what it means to ‘win,'” he said, adding that neither the US nor China can win over the other.He also said the balance can tip too far the other way. Anthropic’s Mythos and Fable models, according to Gates, are tuned so strictly that users asking about cancer end up falling back on the company’s smaller models.
Gates’ warning lands as Trump and AI CEOs sign a voluntary safety pact
The White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities was signed by Anthropic CEO Dario Amodei, Google CEO Sundar Pichai, Meta CEO Mark Zuckerberg, OpenAI President Greg Brockman, Huang and xAI founder Elon Musk. It sets out four layers of checks: internal controls, an internal safety team, independent external auditors and board-level oversight. Trump called it “almost like a constitution,” though the agreement appears to let companies pick their own auditors and cannot be enforced by law.Gates had raised the stakes two days earlier, telling a US television interviewer that AI in the wrong hands is powerful enough to drive events that cause “a billion deaths.”The choice he put to Klein was whether the world waits for a major cyber or bio attack before making safeguards mandatory, or puts them in place before millions die.
Leave a Reply