Speaking at the Financial Times’ Future of AI Summit, the Microsoft co-founder, Bill Gates warned that AI could cause a pandemic deadlier than COVID-19. He criticized the lack of governmental oversight in managing biological threats and cyberattacks arising from the malicious use of AI. He claimed that the Trump administration was blocking critical safeguards against these risks.
Bioweapon Threat
Gates said that an AI-driven biosecurity disaster is likely to be 50 times more scarier than a naturally occurring pandemic. Since frontier AI models significantly lower the technical barriers needed to engineer pathogens, synthesizing viruses has become easier. Historically, synthesing lethal viruses required specialized knowledge. Gates warned that AI could lower some of these barriers. It enables individuals or non-state groups to overcome some of the technical barriers involved.
In addition, a number of developers release models with open-weight access. This allows users to download and run them locally, provided they have suitable computing resources. People seeking to misuse these systems may attempt to modify models to weaken or remove built-in safety filters and monitoring mechanisms of such models. Once these filters are eliminated, the model could provide harmful guidance for dangerous biological agents without sending alerts to the parent company.
Gates emphasized that any model capable of helping design novel molecules should be closely monitored. AI hallucinations can actively designing enitely new variants of deadly pathogens that human immune systems have not previously encountered. He said that an AI-assisted biological incident could result in 15 million excess global deaths compared to COVID-19 between 2020 and 2021.
Infrastructure Attacks and Geopolitical Roadblocks
Rather than relying solely on the expertise of human hackers to detect flaws, advanced AI systems can scan large amounts of software code for security weaknesses. In doing so, they may identify previously unknown vulnerabilities. Once a flaw is detected, some AI systems could help develop and run code to exploit the vulnerability. Gates noted that this poses an immediate threat to critical systems whose failure could have serious consequences. This includes electrical power grids, municipal water networks, and hospital information technology systems.
Gates also criticized the political landscape governing AI regulation. President Donald Trump’s administration has prioritized market-driven self-regulation, favoring voluntary safeguards developed by technology companies. Gates rejects this approach, stressing that voluntary safeguards may prove inadequate when confronted with market competition. When leading corporations are competing for market share, commercial pressures can push safety concerns into the background.
The United States has rejected efforts to develop international AI guidelines, including a G20 initiative, dismissing the G20 effort as a “globalist scheme”. Gates pointed out that because the leading AI laboratories are heavily concentrated in the United States, America’s refusal to participate makes international agreement on safeguards more difficult. If the US remains absent, actions taken by other countries do not bind it.
