Threat of misuse highlights need for both sides to work together on safety: Experts

Shared risks from artificial intelligence are creating incentives for the United States and China to cooperate on safety despite political differences and competition, a Silicon Valley investor and a Stanford University expert say.

The potential misuse of AI in biological weapons development, cyberattacks and other harmful activities presents challenges that could overcome political differences, said Steve Westly, founder and managing partner of The Westly Group, a venture capital firm based in Silicon Valley.

"The issues in AI are going to force a US-China agreement. The sooner we get it, the better," Westly said.

READ MORE: Wang urges China-US cooperation on AI development and governance

His remarks come as the two countries prepare to further engage on AI.

In September, both governments announced a bilateral dialogue on the technology's risks and benefits and agreed to establish a communication channel for AI incidents. China's announcement said the next dialogue would take place in November, while the White House said the next exchange would occur by November.

The commitments provide an opening for cooperation as advances in AI expand both its beneficial applications and its potential for misuse.

The International AI Safety Report 2026 documents increasingly capable systems that can identify software vulnerabilities, generate malicious code and assist with scientific tasks relevant to biological misuse.

In one major cybersecurity competition cited by the report, an AI agent identified 77 percent of vulnerabilities in real software, placing it in the top 5 percent of more than 400"mostly human" teams. The report also notes that it is unclear whether such capabilities benefit attackers or defenders more.

Westly said the scale of the potential dangers would strengthen the case for the two countries to work together.

"There will be challenges with bio-weapons ... AI is going to create some massive challenges," he said. "It will only be solved with the two world superpowers working closely together. It will overcome any conflict and interests we currently have."

For Liang Min, managing director of Stanford University's Bits & Watts Initiative, addressing those dangers does not mean the end of competition, as cooperation and competition can coexist.

Beneficial applications

Min pointed to US-China cooperation on nuclear technology, which can be used for both weapons and electricity generation, as a precedent for managing risks while preserving beneficial applications.

"You may be surprised. The US and China collaborate very closely on nuclear power plant safety," he said.

The two countries will continue to compete regarding who is more advanced in terms of the AI tools and language models, he said, but they should collaborate on AI safety as well, because "that's good for everybody".

The biological risks illustrate the difficulty of separating useful capabilities from potentially harmful ones. In one study cited by the international safety report, one model outperformed 94 percent of domain experts in troubleshooting virology laboratory protocols.

A commentary published by the Brookings Institution in July, which examines how the US and China can cooperate to reduce urgent AI risks, offers proposals for addressing common concerns through regular technical exchanges.

ALSO READ: US urged to collaborate more in the field of AI

The article recommends sharing methods for assessing and mitigating narrowly defined risks, including cybersecurity threats, weapons-related misuse and AI reliability failures that could harm critical infrastructure.

Such exchanges could cover testing before models are released, adversarial testing known as red teaming, and techniques for keeping AI systems within safe environments.

Political mistrust remains a barrier. Westly said the tendency to regard the other country as a threat makes it harder to address common challenges.

While AI development faces energy constraints, he said this was fixable.

"The more challenging issues will be the political ones and the technology ones," he said. "We must find ways to get beyond that."

Contact the writers at liazhu@chinadailyusa.com