
Microsoft co-founder calls for stronger model monitoring and better outbreak detection
AI-generated summary
The article reports an interview by Bill Gates with Madhumita Murgia for the Financial Times podcast Tech Tonic, published on November 4. Gates warns of the risks of malicious use of artificial intelligence in biology and cybersecurity, citing recent research and precautionary measures already taken by developers like Anthropic.
A threat “significantly worse than Covid”. Bill Gates warns of the possible consequences of malicious use of artificial intelligence in biology. In an interview with Madhumita Murgia for the Financial Times' Tech Tonic podcast, published on November 4, the Microsoft co-founder calls on governments to strengthen model monitoring and outbreak detection.
He distinguishes two immediate risks: the use of AI to carry out cyberattacks and its potential assistance in biological weapons projects. According to him, competition between laboratories makes self-regulation insufficient.
Key Points
Bill Gates believes that malicious use of AI could cause a health disaster surpassing that of Covid.
It calls for evaluations of the models and independent audits of their protections.
Researchers have already identified limitations in synthetic DNA order control tools.
He also fears an increase in cyberattacks, particularly those using cryptocurrencies to demand ransoms.
Bill Gates calls for better anticipation of biological risks from AI
Asked about biological risks, Bill Gates raises the possibility of an event “significantly worse than Covid”. It describes a potential scenario, without obviously announcing any identified threat or timetable. His concern concerns the possible damage and the insufficiency of the means of prevention.
In the interview with the Financial Times, he calls for coordinated monitoring of air, wastewater and blood samples. The goal is to spot an infection before it affects a large population.
Developers have already strengthened some protections. Anthropic activated its ASL-3 level for Claude Opus 4 in May 2025, as a precaution, in particular to limit assistance to activities related to biological weapons. This decision did not mean that the company had demonstrated a capability to produce such a weapon.
Furthermore, a study published in Science in October 2025, co-signed by Eric Horvitz, scientific director of Microsoft, showed that sequences redrawn by AI tools could escape the control software of synthetic DNA suppliers. Researchers have developed patches improving their detection. These results are about security filters, not the demonstration of an AI-created pandemic.
Bill Gates also emphasizes the dependence of hospitals, banks and power grids on complex software. In his eyes, advances in models make it easier to find vulnerabilities that could disrupt these infrastructures.
It particularly targets models with accessible settings, from which users can remove protections. He believes that the existence of an economic model based on cryptocurrency ransoms encourages attacks. This observation concerns their criminal use as a means of payment.
Faced with these risks, he calls for public intervention, especially American. It defends independent audits, protection evaluation criteria and work to secure existing software. The founder of Microsoft considers it unlikely that a laboratory alone will slow down sustainably in a competitive market.
AI outlook — possibilities, not facts
Governments, particularly American ones, will announce initiatives to regulate or audit AI models in the coming months.
Likely · Within months
AI model developers will strengthen their protections against assisting in biological weapons activities.
Possible · Within months

Eight days after the $6.43 billion theft from Binance, $3.11 billion remains stuck in addresses linked to the hack, according to Elliptic. Tether and Circle froze $2.85 billion in less than an hour. The FBI attributes the theft to TraderTraitor, a North Korean collective.

Gnosis Pay announced a data breach that occurred on November 3, exposing the names, addresses and wallet data of its customers. The company assures that no funds were stolen and that the affected system will close as planned on December 20.

The XRP Ledger activated amendment fixCleanup3_4_0 on November 4, 2026 to fix a flaw potentially allowing unauthorized token creation via destroy permission. No cases of exploitation have been confirmed.

The Vapor Panda Group uses the ARTEX system, coordinating multiple AI agents, to target more than 200 organizations in 14 countries. This cyberespionage campaign, observed since October, combines automation and human supervision.

Hackers changed the EFS cold room setpoint from −30°C to 110°C via the service provider Coolworld, without loss of pockets thanks to an on-call technician. The attack, which occurred on the night of November 1 to 2 in Burgundy-Franche-Comté, is part of a wave targeting the cold chain in France, Belgium, Ireland and Sweden. ANSSI had alerted about refrigeration controllers accessible from the Internet.

Microsoft is offering a voluntary departure plan to 60,000 employees to reallocate its financial resources towards artificial intelligence infrastructure, despite strong growth in financial results.