OpenAI, Google and Microsoft warn AI could aid biological weapon development, urge new laws

AI company leaders have asked the US Congress to require screening of synthetic DNA and RNA orders. They say the step is needed because AI tools could help people bypass existing biosecurity checks.

In Short

• AI leaders seek safeguards against biological weapon development
• AI can help bypass some genetic screening systems
• Scientists warn existing screening tools are not foolproof

Generative AI has made it easier than ever for people to access information. Users can ask questions in natural language and receive answers that are easy to understand, often within seconds. But that convenience has also raised a troubling question: what happens if someone uses AI to learn how to create biological weapons? That concern is now prompting some of the biggest names in artificial intelligence to call for new safeguards.

AI leaders call for tighter rules

CEOs of several major AI companies are urging members of the US Congress to adopt laws that would make it harder for bad actors to develop biological weapons using AI tools.

Among the signatories are Google DeepMind CEO Demis Hassabis, OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Microsoft AI CEO Mustafa Suleyman. They have signed a public letter calling for laws that would require companies selling synthetic DNA and RNA to screen customers and orders to prevent the misuse of genetic material.

The signatories also include scientists, national security experts, and executives from gene synthesis companies.

How DNA printing became easier

The debate comes decades after scientist Arthur Kornberg first successfully synthesised DNA in the 1950s. Today, the process is largely automated. Dozens of companies around the world use commercial synthesisers to "print" and sell custom genetic sequences that are used in scientific research, drug development, and diagnostics.

Many providers sell only to qualified researchers, biotech companies, and educational institutions. However, not all companies thoroughly vet customers or the genetic sequences they order.

To reduce risks, many providers already use software that screens orders for "sequences of concern" — genetic material that could contribute to toxicity or help cause disease.

Where AI becomes a concern

According to the concerns raised by the signatories, AI tools can make it much easier for someone to identify suppliers that do not screen orders.

If prompted appropriately, AI systems can also suggest ways to modify an order so that even companies that conduct screening may find it more difficult to detect what a customer is attempting to create.

Researchers also worry that AI could make it easier to design dangerous new toxins and pathogens. While some biology expertise would still likely be required to create a functional virus from scratch, AI could lower some of the barriers that previously made such work more difficult.

Existing safeguards have limits

The concerns are not purely theoretical. Last year, Microsoft researchers published a study showing that AI-powered protein design tools were able to generate potentially dangerous gene sequences that slipped past screening software used by some companies.

The models proposed new protein sequences that shared structural similarities with known dangerous proteins, highlighting potential gaps in existing safety systems.

- Ends