Bill Gates is calling for a fundamental change in how governments approach artificial intelligence: treat powerful AI more like medicine, aviation or automobiles—technologies where innovation is permitted, but safety cannot be left entirely to the companies building the product.
In recent interviews, the Microsoft co-founder has sharply escalated his warnings about AI. Gates told NBC’s Meet the Press that AI could be powerful enough, if misused, to contribute to events causing catastrophic loss of life. In a separate conversation on The Ezra Klein Show, he argued that relying solely on the AI industry to regulate itself is inadequate, pointing to the existence of independent safety requirements in pharmaceuticals, aviation and automobiles.
Gates Compares AI With Medicines
The pharmaceutical analogy captures the heart of his argument. Society does not allow a drug manufacturer to independently decide that a medicine is safe and release it without external oversight. Medicines go through testing and approval, followed by monitoring once they enter widespread use.
Gates believes advanced AI needs an analogous governance architecture—not necessarily identical regulation, but independent evaluation, minimum safeguards, continuous monitoring and accountability.
The comparison extends naturally to aviation and automobiles. Aircraft manufacturers operate under common safety standards; vehicles must satisfy defined requirements. Gates argues that technologies capable of creating systemic harm warrant safeguards that do not depend simply on which company behaves most responsibly.
Cyber and Bio Risks Change the Equation
Gates' concern is increasingly centred on AI's dual-use capabilities.
In cybersecurity, advanced models can help defenders discover vulnerabilities, analyse malware and accelerate remediation. But similar capabilities can potentially help malicious actors discover weaknesses or scale attacks.
Biological misuse worries him even more. Gates has argued that AI could lower barriers to capabilities that historically demanded substantial expertise and resources, potentially enabling smaller groups to pursue dangerous biological threats.
This creates an uncomfortable paradox: the more capable AI becomes at solving complex scientific and security problems, the more valuable those same capabilities may become to attackers.
Beyond the AI ‘Kill Switch’
Gates also challenges the popular notion that AI safety can ultimately be reduced to an emergency off switch.
His argument is that the immediate challenge is not simply whether humans can turn a computer off. It is whether organizations can detect dangerous activity while AI is operating.
That shifts attention toward monitoring, authorization, access control, auditability, containment and continuous risk assessment. Gates has specifically emphasized safeguards capable of identifying potential cyberattack and bioterrorism misuse.
In other words, AI safety increasingly has to operate at runtime, not merely at model release.
The Drug Analogy Has Limits
There is, however, an important difference between AI and pharmaceuticals.
A medicine is developed for comparatively defined purposes and can undergo controlled clinical trials. AI is a general-purpose technology. The same foundation model can write software, analyse medical research, generate marketing content, operate an autonomous agent or support cybersecurity work.
Models are also continuously updated and can be combined with external tools, proprietary data and autonomous agents.
That makes a conventional one-time regulatory approval difficult. AI may require something closer to continuous certification—evaluating not only the underlying model but also how it is deployed, what data it accesses, which tools it controls and what actions it is permitted to execute.
Regulation also brings trade-offs. Excessive compliance requirements could raise costs, slow beneficial applications and disproportionately advantage large companies capable of absorbing regulatory overhead. Monitoring open-source AI introduces additional questions around privacy, decentralization and enforceability.
From ‘Ship Fast’ to ‘Prove Safe’
The deeper significance of Gates' argument is therefore not that AI should literally be regulated exactly like a pharmaceutical drug.
It is the principle behind the analogy: powerful technology should have evidence-based safeguards proportionate to the damage it could cause.
The AI industry is rapidly progressing from chatbots toward autonomous systems capable of reasoning, accessing tools and executing actions. Governance consequently must evolve from static policies toward continuous verification and oversight.
AI should function like a seat belt—an essential protection layer, but never a substitute for responsible human judgment, governance and accountability.
The technology race is therefore acquiring a second finish line. Alongside the competition over who builds the most powerful model or ships innovation fastest is an equally consequential contest:
Who can earn trust first?
See What’s Next in Tech With the Fast Forward Newsletter
Tweets From @varindiamag
Nothing to see here - yet
When they Tweet, their Tweets will show up here.
