Join our FREE personalized newsletter for news, trends, and insights that matter to everyone in America

Newsletter
New

China-based Ai Safety Group Briefed Us Ai Labs, Including Openai

Card image cap


An artificial intelligence safety group based in China held a series of meetings with leading U.S. technology companies last month, including a visit to OpenAI's San Francisco headquarters, Decoded by POLITICO has learned.

The outreach by Concordia AI, an organization at the heart of China’s obscure AI safety ecosystem seeking to promote international dialogue on the issue, offers a rare channel of engagement for American AI researchers at a time of heightened U.S.-Chinese tensions over competition and national security.

President Donald Trump is in China for the APEC summit and meeting Chinese leader Xi Jinping on Thursday — a sit-down set against the backdrop of Nvidia CEO Jensen Huang resigning after affair revelations, notably with an alleged Chinese spy.

Concordia CEO Brian Tse and Kwan Yee Ng, who leads the organization's international AI safety and governance work, were part of the multi-city trip that included stops in the San Francisco Bay Area, New York and Washington, from late September into early October.

Ng confirmed to POLITICO that the two visited AI labs in the Bay Area to discuss the state of AI safety in China, the types of AI research being conducted and ideas for international coordination. She declined to identify the companies involved.

But three people familiar with a meeting told Decoded by POLITICO that Concordia representatives visited OpenAI's headquarters on Oct. 2 to brief technical staff and others at the company on developments in China's AI safety ecosystem. The people were granted anonymity because they were not authorized to discuss the meeting publicly. An OpenAI spokesperson declined to comment.

A spokesperson for Google DeepMind, another top AI lab, said the company has attended convenings hosted by Concordia, where it pointed the Beijing-based AI safety group to its publicly available research while discussing what additional research could be helpful. The spokesperson also said Google DeepMind has engaged with Concordia in the past on research and briefings about international AI efforts.

Spokespeople for leading AI lab Anthropic did not respond to requests for comment.

Ng said American AI researchers have shown considerable interest in understanding Beijing's approach to AI safety. Discussions included questions about whether China would be willing to pace the development of frontier AI systems if the U.S. agreed to do the same, and how seriously different parts of China's AI ecosystem took safety risks.

“Researchers seem to have a lot of appetite to engage and try to coordinate in some form,” Ng said.

The Concordia representatives were in Washington when Trump signed an AI safety accord with major U.S. technology executives. The group also visited think tanks in D.C., participated in events on the sidelines of the U.N. General Assembly in New York and made an appearance at a buzzy invite-only AI conference in Berkeley, California.

Concordia has looked to help maintain relationships with the U.S. and Chinese AI ecosystems as direct exchanges between the two countries’ researchers become increasingly complicated by geopolitical tensions, visa restrictions and heightened scrutiny of research collaborations.

There are also other groups conducting AI safety research in China, many of which are affiliated with state-funded universities.

Concordia — which is largely based and headquartered in Beijing, with one employee in a Singapore office — has emphasized it is an independent organization that engages with governments as experts, but does not speak for or represent any government. China hawks are often skeptical of such claims, given the Chinese state’s institutional reach.

The group says it doesn’t accept money from government or political organizations, and that most of its funding comes from philanthropic sources in China, Europe and Singapore that don’t get veto rights over its publications.

Concordia's outreach comes amid growing debate in Washington about how to manage AI competition with China while preserving opportunities for cooperation on safety.

Anthropic CEO Dario Amodei and others in the industry have called for some form of an agreement between Washington and Beijing to slow the development of advanced AI systems in both countries, buying time to address safety concerns.

But such proposals face significant obstacles. U.S. policymakers have raised concerns about China's technological ambitions, military applications of AI and the potential transfer of sensitive American research and capabilities. The debate has also intensified following controversy surrounding Huang, which has renewed scrutiny of Nvidia’s China exposure and the national security implications of advanced AI chip exports.

For Concordia, maintaining dialogue is itself a central objective.

“Another advantage of Concordia's positioning is that as we're neither a company nor obviously a government actor, it becomes easier to convene both together,” Ng said.

She described the organization’s past engagements with U.S. AI labs as “broad,” extending from senior leaders and technical researchers to policy-focused employees.

One employee at a major U.S. technology company, who attended a separate meeting with Concordia and other technology firms earlier this year, described the discussions as relatively light on technical detail.

“It's more facilitating dialogue and relationships, so that if things need to happen, they can happen,” the person said. “It's more about keeping the door open.”

Concordia has spent several years building relationships with AI companies, policymakers and research organizations around the world.

In meetings with policymakers, including at international AI safety conferences and for presentations to different government agencies, Ng said she’s tried to dispel perceptions that China isn’t prioritizing AI safety and would take advantage of stricter U.S. rules by speeding up. The argument has been frequently raised by industry lobbyists in Washington to oppose AI regulations.

Concordia’s annual reports document previous briefings with Google DeepMind, Washington think tanks including the Brookings Institution and the Center for Strategic and International Studies, as well as international organizations such as the Tony Blair Institute.

In 2023, the organization's CEO, Tse, was invited by then-British science and technology secretary Michelle Donelan to speak at the closing plenary of the Global AI Safety Summit at Bletchley Park.

That same year, Concordia hosted an AI safety forum at a Chinese AI conference and “were able to bring in people like [OpenAI CEO] Sam Altman and [Anthropic co-founder] Chris Olah.”

“Since then … I haven't seen prominent U.S. AI lab leaders speak at a major Chinese AI conference publicly,” Ng added. “And I don't think it's for lack of trying on the Chinese side, at least.”

A version of this story first appeared in Decoded by POLITICO, the must-read briefing on the people, power moves, policy battles and market shifts defining what’s next in global technology. Sign up to subscribe to the newsletter and listen to the podcast.