In early November, a half dozen of America’s most senior retired military officers, including former NATO Supreme Allied Commander Europe, Curtis M. Scaparrotti, and the former director of the National Security Agency, Michael S. Rogers, quietly met in Beijing with retired Chinese officers of similar rank. With President Biden scheduled to meet a week later with Chinese leader Xi Jinping in San Francisco, Pentagon and White House officials asked the Americans to probe whether China seemed ready to resume military communications after 16 months of refusing to pick up the phone.
Often in these sorts of meetings, China trots out generals who are known more for their political skills in dealing with foreigners than their military chops — “barbarian handlers,” as they are dismissively called. But in this meeting, organized by the Stimson Center, a Washington D.C. think tank, the Chinese side was stocked with former commanders who had deep operational experience and had wrestled with the question of what might happen if ships or planes of the two nations collided.
During the talks, the Chinese relayed two messages: China was serious about resuming contacts but that didn’t mean it accepted U.S. patrols of the South China Sea and Taiwan Strait.
“The Chinese said there could be interest in resuming mil-mil communications but emphasized preconditions more than resumption,” says a person familiar with the talks. “The Chinese are using brinkmanship to trade for the frequency of U.S. military surveillance in the South China Sea. Their message is ‘If you don’t want a crisis, patrol less.’”
That information fed into the preparation for the summit and helped ease negotiations over the issue. In a press conference after his meeting with Xi, President Biden announced that the U.S. was “back to direct, open, clear, direct communications.” A month later, the Pentagon’s top officer talked by videoconference with his Chinese counterpart.
U.S. officials believe many more government negotiation sessions may be required before such talks become routine, especially with an ongoing shakeup in China’s military. But the previously unreported military discussions in Beijing were a triumph of what’s known as “track-two” diplomacy — sessions involving retired government officials, business leaders, academics and think tank experts from different nations who try to work out the knotty problems dividing their governments.
We follow these sessions incredibly closely… We try to get a sense of the areas of focus for the Chinese. What is top of mind? Is there any maneuvering room?
A senior U.S. administration official
Because the talks are unofficial and almost always off-the-record, the participants can talk more candidly than government officials and spitball different ideas that might be proposed in official, “track-one” sessions.1Some analysts also talk of “track-1.5 sessions” where current government officials take a bigger role. We included those talks in our discussion of track two because so many track-two discussions also involve meetings with current office holders. All sides understand that track-two participants report the conversations to their government contacts, who then use the information to size up the opposing capital and help formulate policy.
Track-two sessions have sometimes played starring roles in post-World War II diplomacy. After an American U-2 spy plane was shot down over Soviet airspace in 1960, Norman Cousins, the editor of the Saturday Review, diffused tensions by organizing an unofficial U.S.-Soviet meeting at Dartmouth College. Track-two negotiations also paved the way for the 1993 Oslo Accords, a peace deal between Israel and the Palestine Liberation Organization, and helped lift roadblocks to the 1998 Good Friday Agreement, which ended sectarian violence in Northern Ireland.
Unofficial talks also nurtured U.S.-China relations, starting when the National Committee on U.S.-China Relations, a pro-engagement group, hosted Chinese ping pong players on a 1972 goodwill visit to the United States. Track-two sessions also acted as an unofficial channel when relations between the two governments froze after China’s massacre of students in Tiananmen Square in 1989 and the U.S. bombing of China’s embassy in Belgrade 10 years later.
After declining during the Trump years and the height of the pandemic, track-two dialogues have more recently revived — a sign, says a senior U.S. official, that Beijing and Washington are serious about stabilizing relations.
Currently there are at least 40 different U.S.-China track-two groups. The majority focus on the economy and national security, though some discuss the environment, healthcare, technology and human rights. U.S. officials say the Biden administration is most interested in sessions involving military-to-military communications, Taiwan, the South China Sea and the economy — and especially who in China has Xi Jinping’s ear on these issues.
U.S. PARTICIPANT | CHINESE PARTICIPANT | SUBJECT |
---|---|---|
National Committee on U.S.-China Relations | China Foundation for Human Rights Development | Human rights and rule of law |
China-U.S. Green Fund | Digital economy and AI | |
Peking University’s National School of Development | Healthcare | |
Tsinghua Universtiy’s Institute of Energy, Environment and Economy | Climate finance | |
Chinese economists | Economics and trade | |
National Institute for South China Sea Studies | Martime issues and international law | |
Chinese security experts | National security | |
U.S. Chamber of Commerce | China Center for International Economic Exchanges | Business and economics |
U.S.-China Business Council | Chinese economic officials | Business and economics |
Center for Strategic and International Studies | Shanghai Institute of International Studies | National security and economics |
Peking University’s Institute of International and Strategic Studies | Scholarly exchanges | |
Brookings Insitution | Tsinghua University’s Center for International Security and Strategy | Military uses of AI |
INHR | Chinese Arms Control and Disarmament Association | Military uses of AI |
Centre for Humanitarian Dialogue — Geneva | Tsinghua University’s Center for International Security and Strategy | Military uses of AI |
National Committee on American Foreign Policy | Chinese and Taiwanese experts | U.S.-Taiwan-China relations |
Quincy Institute | China Foundation for International and Strategy Studies | Crisis management |
Stimson Center | Chinese military officials | Military relations |
George H.W. Bush Foundation for U.S.-China Relations | Peking University’s Institute for Global Cooperation and Understanding | National security |
Aspen Institute and World Resources Institue | Counselors Office of the State Council of China | Climate change |
Bonnie Glaser, German Marshall Fund | Chinese experts | Taiwan Strait issues |
Source: The Wire China research, RAND Corp.
Before Saturday’s Taiwan election, for instance, some track-two groups relayed to Beijing the steps the U.S. government might take in congratulating the new president, whoever was elected, as a way to assure China that an uptick in U.S.-Taiwanese meetings didn’t mean Washington was trying to encourage Taiwanese independence.
“We follow these sessions incredibly closely,” says the senior administration official. “We always debrief intensively a number of people in the groups. We try to get a sense of the areas of focus for the Chinese. What is top of mind? Is there any maneuvering room?”
On the Chinese side, some participants work for think tanks funded by the Ministry of Foreign Affairs and the People’s Liberation Army. Others are in close contact with those agencies as well as others.
“Officials at the Ministry of Foreign Affairs know 20 or 30 scholars who have more communications with American colleagues,” says Da Wei, director of Tsinghua University’s Center for International Security and Strategy. “Government officials may want to listen to us.”
U.S. and Chinese participants in track-two sessions have objectives and motivations ranging from individual aggrandizement — free trips, schmoozing with senior government officials, networking for business — to exploration of issues that hamstring the governments. Sometimes official talks are stuck because the politics are tough, such as restarting military talks. Other times it’s because the issues are so fresh and complex, such as making sure artificial intelligence doesn’t turn into a Terminator technology.
Track-two sessions can provide intelligence on what the other side is thinking and suggest policy proposals but they can’t substitute for government action. Still, the fact that these kinds of talks are picking up after years of deteriorating relations between Washington and Beijing suggests efforts in both capitals to stabilize the world’s most important bilateral relationship.
AI ARMS CONTROL
When it comes to U.S.-China relations, Henry Kissinger made history in his track-one talks with Mao Zedong and Zhou Enlai, and later became a master at using track-two sessions for business and policy. From 2009 until about 2013, the former Secretary of State organized one of the most high-profile track-two dialogues, what the Chinese named the “China-U.S. Track Two High-Level Dialogue.” Kissinger gathered such retired U.S. luminaries as Secretary of State George Schultz, Secretary of Treasury Robert Rubin and Secretary of Defense William Perry to the cause, while the Chinese delegation was led by Tang Jiaxuan, once China’s most senior diplomat, who arranged for the group to meet with China’s senior leaders, including Xi Jinping.
“I think Henry’s idea was just to engage with Xi Jinping and Putin in Russia in the hopes that would help further better understanding,” says Rubin. (Kissinger organized similar trips to Moscow.) The meetings also helped Kissinger’s consulting business. They were covered by Xinhua, China’s state-run news agency, and helped remind his clients of his China clout.
Before he died in November, the 100-year-old Kissinger most recently used track-two sessions to lobby Beijing on a particularly futuristic problem set: limiting the harm that could be caused by artificial intelligence, especially if the software is used to fire weapons.
Kissinger traveled to Beijing in July to meet with Xi Jinping. For Beijing, Kissinger’s trip was a way to tweak the Biden administration about its hard-line China policies by feting him as an “old friend,” while John Kerry, Biden’s climate envoy who was also in Beijing at the time, couldn’t get a meeting with Xi. For Kissinger, it was a chance to press Xi and other senior officials in-person about how the two nations could control AI, says Harvard’s Graham Allison who co-wrote a Foreign Affairs article with Kissinger in October. Their plan includes setting up track-two talks between “the two AI superpowers” to quash applications “that pose catastrophic risks.”
Track-two talks play an especially important role on issues where the capitals haven’t yet established policy and are more open to suggestions from outsiders. In the 1990s, dialogues on Taiwan policy helped Beijing and Taipei work out a way to increase economic ties while putting off questions about reunification. In the 2000s, economic talks on Chinese barriers to foreign investment helped shape trade negotiations. Now, both countries have begun to focus track-two energies on AI.
Apart from Kissinger’s push for joint U.S.-China talks, at least three track-two dialogues on the military use of AI have been meeting quietly for five years. These participants see their discussions on AI as a 21st century version of nuclear arms control.
There’s a growing, international community of AI management experts. This is not only between China and the U.S., but between scholars and think tanks from many fields.
Qi Haotian, Deputy Director of Peking University’s Center for International Security and Peace Studies
In the 1980s, the Soviet Union and the U.S. had a near monopoly on nuclear weapons and the underlying technology, which could do unimaginable damage if unleashed. While China and the U.S. can’t claim that degree of control over AI research, they are the leading powers in a race to create a technology that could reshape the world in ways that even those building the systems can’t predict.
But there are important differences. While limits on nuclear weapons could be verified by inspecting missile silos, there is no way currently to verify limits on AI. Instead, experts are pushing for agreements on best practices. And while the architects of nuclear weapons worked for the U.S. or Soviet governments, leading AI experts work outside the government, giving added impetus for track-two talks.
Qi Haotian, deputy director of Peking University’s Center for International Security and Peace Studies, says track-two talks can help build an influential group in China that sees the value of avoiding an AI arms race — eventually easing opposition in Beijing.
“There’s a growing, international community of AI management experts,” he says. “This is not only between China and the U.S., but between scholars and think tanks from many fields.”
The AI track-two talks have proceeded down two paths, which often overlap. One set of discussions, organized by the Brookings Institution and Tsinghua’s CISS, has met eight times since 2019 and focuses on broader issues, including devising common definitions of the systems being discussed. Unless Chinese and American negotiators can agree on what an “autonomous weapons system” is, for instance, it would be impossible to decide how to control it.
“It was hard to find a similar term” in Chinese, says Xiao Qian, CISS’s deputy director, who scoured academic papers in the West and China for help. So far, the two sides have agreed on a glossary of about two dozen definitions.
On the Chinese side, those talks have been led by Fu Ying, a former ambassador and vice minister of foreign affairs who has deep contacts in the Chinese military and foreign affairs establishments.2For a time, comments attributed to Fu on social media were blocked because they were seen as obliquely critical of Xi’s focus on a rising China vs. a declining West. But now she is once again able to publish. Some in the U.S. government take that as a signal that Fu has once again become influential. But Fu’s colleagues say she didn’t make the comments and they removed as a way to correct the record. And on the U.S. side, they were initially led by Brookings chief John Allen, a retired Marine Corps general. After Allen resigned from Brookings, he was replaced by Colin Kahl, a former Biden undersecretary of defense. Their stature has helped elevate the track two’s work in both capitals, although much of the work was done by Brookings China specialist Ryan Hass and AI expert Chris Meserole, now at the Frontier Model Forum, an AI policy group.
“During track-two discussions with our colleagues from the U.S., we focused on how to define the boundaries of ‘off-limits areas’ for AI-enabled weapons,” Fu wrote in 2020.
A second set of talks drafted a proposed Code of Conduct where the two governments would agree to prohibit AI programs from firing weapons systems without a human giving the order and would keep AI programs far from the chain of command needed to launch nuclear weapons. The talks initially were held under the auspices of the Centre for Humanitarian Dialogue, a mediation group in Geneva, which is now trying to figure out how to get other countries interested in adopting the Code.
A third group split off from the Centre for Humanitarian Dialogue and meets nearly monthly trying to suss out how limits on AI would work. The Pentagon, State Department and Chinese Ministry of Foreign Affairs sometimes send observers to watch these discussions firsthand.
On the American side, the talks have been led by a group including former State Department lawyer Eric Richardson, retired Lt. Gen. Jack Shanahan (who was the first director of the Pentagon’s Joint Artificial Intelligence Center), and Paul Scharre, a former Pentagon drone expert who is now at the Center for a New American Security, a Washington think tank. The Chinese side is represented by former foreign affairs officials now at the Chinese Arms Control and Disarmament Association, a Chinese think tank, and former PLA officials at the National Defense University and National University of Defense Technology.
Topics under consideration include coming up with ways to test AI systems before they are deployed, convincing smaller states to adopt similar rules, deciding whether AI should be used in the early-warning systems that feed into nuclear weapons, and dealing with AI-weapons systems that misfire and accidently kill civilians or troops. One idea the group has discussed, says Shanahan, is labeling AI weapons systems so adversaries know what they are facing. Another is to create an international organization that investigates alleged misfires.
“If an AI system shoots a missile you never wanted it to shoot, then what do you do?” says Shanahan. “Does the other side believe it’s an accident? Throw AI into the equation and it can escalate very quickly” through automated retaliation.
To help work through these problems, the group produces video simulations of incidents involving artificial intelligence that could lead to war—though the combatants are called “Redland” and “Blueland” rather than the U.S. and China to make them less emotionally charged. In one video, a Blueland drone fires a missile that mistakenly hits a Red Cross convoy and kills dozens of Redland soldiers. Although the Blueland president then orders a ban on AI weapons systems, Redland starts using its AI weapons to turn back Blueland’s advance as the conflict escalates.
At the end of the video, Richardson, playing the role of a newscaster, asks, “How could this accident have been prevented and the risks of using artificial intelligence mitigated?”
The video set up the discussion among the 10-20 track-two participants, who are split into mixed teams of Americans and Chinese. “We have done probably 20 scenarios over five years,” says Richardson. “All reveal practical new tips.”
The work from these AI-military discussions fed into the preparation for the Biden-Xi summit in San Francisco — but it’s not clear it will do much to sway the Chinese side. The Pentagon has already adopted a policy of having commanders “exercise appropriate levels of human judgment” before AI-enabled weapons are fired, but the Chinese government hasn’t. During his discussions with Xi, Biden raised U.S. fears about mixing AI and nuclear weapons, says the senior U.S. official. While Xi indicated he understood the importance of AI technology, he was noncommittal about its use.
During the summit, the two governments committed to talks on artificial intelligence, though the agenda is still being worked out. Some in the White House fear China will use the talks to learn about advanced U.S. technology, says the senior administration official.
China’s wariness is even greater, fearing that any discussion of AI control — especially prohibiting the use of AI in nuclear launch systems — would be used by the U.S. to try to contain China’s nuclear buildup. The Soviet Union fell into an American trap by limiting its nuclear arsenal via arms control, according to that view, and China won’t make the same mistake. China’s nuclear establishment is “an isolated kingdom,” say national security experts, and especially resistant to change. Even the names of some of the agencies involved are considered secret.
REAL WORLD RESULTS?
Measuring the success or impact of track-two talks is not a straightforward process. Government policies are affected by many different factors, after all; some talks feed into concrete policies and actions, while others are simply ways to gather information on what the other side is thinking.
“The main goal is to clarify for both sides the variety of points of view in both countries and reduce misperceptions,” says Scott Kennedy, a Chinese economy expert at the Center for Strategic and International Studies, who organized sessions last year on scholarly exchanges along with Peking University’s Institute of International and Strategic Studies.
Simply clearing up misperceptions can be an important outcome. Since around 2004, one dialogue — led by Michael Swaine at the Quincy Institute, a Washington think tank, Iain Johnston, a Harvard China expert, and the China Foundation for International and Strategy Studies, a PLA-funded think tank — has focused on crisis management. In recent sessions, says Peking University’s Yu Tiejun, the group has worked through scenarios involving crises in the South China Sea and Taiwan Strait. Yu says the war gaming has made him realize that both sides had different understandings of international law and what was considered a hostile action.
“The new generation knows very little” about U.S.-China crisis management, says Yu. And with considerable turnover in government, the dialogues serve a further purpose: “to retrain people in important positions.”
The frank communication style and unique political access that are found in track twos are often cited as advantages, but critics like to point out that the process can be lopsided. In reality, Chinese participation can be very limited.
The sessions are mainly conducted in English, for instance, ruling out anyone who doesn’t have excellent language skills. And on the economic front, the Chinese participants almost uniformly believe in opening their markets to the West. Rarely, if ever, are neo-Maoists who reject that view included in the talks although they may be influential in Beijing.
Even so, Da, the Tsinghua professor, says the talks still have payoffs, but “it’s more or less indirect.” For several years, he has been participating in track-two talks about increasing flights, tourism and student exchanges between the U.S. and China, and he reports the results to government officials overseeing foreign affairs, education, the economy and propaganda.
“Track-twos build consensus first among scholars and then among those in decision-making circles,” he says. With what he calls people-to-people exchanges, “the government lags behind the track-twos by three to six months.”
As evidence that consensus-building works, Da points to Xi’s comments in San Francisco where he invited 50,000 American youths to study in China over the next five years. More sensitive topics, however, like boosting the number of journalist visas or reopening consulates closed during the Trump years, remain off limits, he says.
But others, especially those in the Trump administration, say that track-two talks are a waste of energy.
“I don’t think they serve much of a purpose other than padding the airline mileage accounts of the participants,” says Matt Turpin, the Trump National Security Council’s China director and now a visiting fellow at the Hoover Institution at Stanford. Nothing in track-two briefings, he says, “induced us to rethink what we were doing.”
Turpin believes there is an insidious side to the discussions too. When Chinese participants parrot the Communist party’s message in their talks with American colleagues, those views make it into op-eds and other writings published in the United States. But American participants don’t have the same way to influence public opinion in the closed Chinese system.
Still, Turpin is one of the few former Trump senior officials to join track-two talks. “Since these go on, it’s important that Beijing has exposure to a variety of views,” he says, not just pro-engagement messages.
When I add up track-two sessions, do I think they influence the Chinese leadership? Yes, I think they do.
A U.S. Chamber of Commerce Track-Two dialogue participant
For some track-two efforts, business — not government policy — is the objective.
The U.S. Chamber of Commerce, for instance, started annual track-two sessions in 2011 with the China Center for International Economic Exchanges, a government-funded think tank. The Chamber dialogue differs from most because participants are mainly chief executive officers of prominent companies in both countries, not economists or consultants. The Chamber first tried to use the talks to build support for a free trade agreement with China and then an investment treaty. Neither proposal made serious headway in government circles, but the efforts kept trade issues at the top of U.S-China agenda, particularly during the Trump years when the U.S. and China hammered out a trade agreement.
“We provided to the Chinese our thoughts on issues that would be incorporated in part by the Trump administration in the Phase One trade agreement” as a way to get China to take the issues more seriously, says Myron Brilliant, the Chamber’s former executive vice president.
For CEOs, the Chamber talks have become a way to lobby Beijing, not necessarily by bringing up specific regulatory problems but by demonstrating that they are working to steady the U.S.-China relationship. They, like Kissinger, hope to be treated as ‘old friends.’
“Given deepening bilateral tensions, it’s never been more important for business, former senior officials and leading scholars” in both countries to be able to talk candidly, says a Chamber spokesman.
The Chamber doesn’t publish the names of its track-two participants, but at the last meeting in Beijing, in July 2023, the Chamber delegation included CEOs from semiconductor makers Intel Corp. and Qualcomm Inc., which get a big chunk of their revenue from China sales, and others from the financial services, technology and energy fields. They met with China’s vice president Han Zheng, senior officials in the ministries of commerce, technology and development, the former longtime Chinese ambassador to the U.S. Cui Tiankai, and Liu He, who was Xi Jinping’s economic envoy to the Trump administration and is now expected to play a big role in track-two sessions.
“These are pretty heavy hitters” in an opaque system, says one participant. “When I add up track-two sessions, do I think they influence the Chinese leadership? Yes, I think they do.”
Given the suspicion in both capitals, however, the question of influence has become thornier. The very same track-two effort that helped unfreeze military-to-military communications last year also has a checkered history as one of the most controversial track-two efforts.
The talks between retired 4-star military officers, launched in 2008 by retired Adm. William Owens, a former vice chairman of the Joint Chiefs of Staff, were dubbed the Sanya Initiative, for the Chinese vacation spot where early discussions were held. Owens worked with the China Association for International Friendly Contact, a PLA organization, to arrange the annual meetings.
Back then, when the U.S. and China were both pursuing a policy of engagement, Owens became a Huawei consultant and later chairman of a firm that teamed with Huawei to try to win telecommunications contracts in the United States. In 2009, he also called for improving relations with China by cutting off arms sales to Taiwan. As concern grew that Huawei was helping Beijing spy on clients, Owens became a target of attacks in conservative media including Fox News.
The Sanya Initiative was partially funded by insurance executive Maurice “Hank” Greenberg, one of the most successful American businessmen in China, as well as the China-U.S. Exchange Foundation, a Hong Kong group that promotes engagement but which the Jamestown Foundation, a defense think tank in the U.S, says has deep connections to the Chinese Communist Party and its overseas “united front” operations. CUSEF didn’t respond to several requests for comment.
During the Obama administration, Sanya Initiative generals met regularly with U.S. officials. Secretary of State Hillary Clinton even held a dinner for visiting retired Chinese generals and their American counterparts, according to Owens.
But reception for Sanya initiatives grew chillier during the Trump administration, and the group moved to the Stimson Center in 2020 with Owens taking a much-reduced role, partly as a way to win the confidence of the Pentagon. Owens, who is now 83, didn’t sit in on the 2023 talks in Beijing, for instance. Sanya now gets its funding entirely from U.S. sources, not CUSEF, says Yun Sun, director of the China program at the Center, but its recent success suggests there is still a path forward for U.S.-China relations.
“We reinvented it,” she says. “And in Beijing the most acute and critical concerns were discussed.”
Bob Davis, a former correspondent at The Wall Street Journal, covered U.S.-China relations beginning in the 1990s. He co-authored “Superpower Showdown,” with Lingling Wei, which chronicles the two nations’ economic and trade rivalry. He can be reached via bobdavisreports.com.