Dr. Matthew Botvinick, a neuroscientist and AI researcher on Anthropic’s technical staff, warned this week that artificial intelligence could deepen threats to constitutional democracy by amplifying social insecurity when trust in government is low. His comments came in an interview with Mercy A. Kuo, published October 7 by The Diplomat, alongside two other analyses of an Asia-Pacific security order in flux: a new three-nation defense pact in the Muslim world and Taiwan’s rethink of how it would fight.
Democracy’s Exposure to AI
Botvinick, author of "AI and Political Reform: The Risk to Democracy and How to Respond" (Princeton University Press 2026) and a senior fellow at Yale Law School, said insecurity combined with low institutional trust creates conditions for what he called "the demagogue’s playbook." In it, leaders exploit populist sentiment to undermine democracy. Kuo said this was the 529th interview in "The Trans-Pacific View Insight Series."
He stressed that these were risks, not certainties. AI could also help democracy, he said, if institutions reform emergency law, surveillance law, separation of powers and free-speech doctrines. Over the longer term, it might inform citizens, support democratic deliberation and improve government transparency and accountability.
He described democracy as a balance between government power and society’s power. Dividing authority among institutions checks government. Society’s capacity for collective action rests on news media, civil society, a free marketplace of ideas, and free and fair elections. AI could tilt that balance, he warned, by over-empowering the executive when leaders use it to automate obedience to central authority, or by weakening citizens’ ability to act together through AI-enhanced surveillance or propaganda.
China offered a warning, Botvinick said. AI can analyze vast data produced by cheap, ubiquitous sensors like cameras and by online activity, helping an authoritarian government monitor and control people. In democracies, the challenge is meeting legitimate law-enforcement and national-security needs without building tools that are easy to abuse. He cited a record of U.S. abuses, from FDR and J. Edgar Hoover to Edward Snowden’s disclosures, and said civil liberties and the rule of law should be protected before problems arise.
Constitutional democracy, he said, is "absolutely mandatory for AI safety." In an authoritarian state, even a technically safe AI system could help leaders crush dissent and enforce compliance. He named China, Venezuela, Russia and some Gulf states. Speaking in a personal capacity, not for Anthropic, he said AI companies, including all major labs, have a responsibility to use their power to protect and strengthen democracy at home and abroad. Democratic backsliding in recent decades, he noted, has more often involved governments subjugating corporate power than the reverse.
A Pact With Unanswered Questions
In a separate analysis, Natiq Malikzada examined the Mecca Joint Defense Agreement, signed August 7, 2026, by Turkiye’s Recep Tayyip Erdoğan, Saudi Arabia’s Mohammed bin Salman and Pakistan’s Shehbaz Sharif. It commits the three to collective defense. On August 31, their foreign ministers, defense ministers and military chiefs met in Istanbul and announced a secretariat in Saudi Arabia, initially headed by a Pakistani secretary general.
The full text hasn’t been made public. A Pakistan Foreign Ministry statement said an attack on any of the three would be treated as an attack on all. But Malikzada said the announced terms don’t specify which forces would fight, whose command they’d follow, or what timetable they’d use. Nor do they explain how members would decide military action was necessary, share costs or define common threats.
The economics are lopsided. Saudi Arabia has oil wealth, Turkiye has industrial capacity but struggles with inflation, and Pakistan is an IMF borrower. The accord has no equivalent to NATO’s formula-based common funding. Malikzada asked whether Riyadh would demand influence over deployments if it paid most bills, and whether Turkiye and Pakistan would accept that while supplying forces and bearing casualties.
The strategic picture is just as uneven. Pakistan didn’t join in defending Saudi Arabia against current attacks by Iran, and Pakistan and Turkiye both fear being drawn in after the Houthis threatened they would pay a "high price" for intervening. Oil loadings at Yanbu have stopped after drone attacks, and experts said the pause could further disrupt global energy markets. SIPRI data show China supplied 80 percent of Pakistan’s major arms imports in 2021–2025, while the United States supplied 77 percent of Saudi Arabia’s, complicating any integration of forces.
Taiwan Trades Plans for Endurance
In a third piece, K. Tristan Tang and H. Rion Wu wrote that Taiwan has shifted from seeking a decisive battle toward an erosion strategy and sustained operations. More than one-third of the population lives in the north, home to the political center and key command and transport hubs. During exercises, nearly two-thirds of active-duty ground forces could be positioned there.
The 2026 calendar included a five-day Immediate Combat Readiness Exercise in June, a five-day Joint Defense Exercise in July, a 10-day Han Kuang in August and a 10-day Lu Sheng expected in October. Maj. Gen. Lin Shih-wei said Lu Sheng focused on successive resistance and sustained operations under Chief of the General Staff Admiral Mei Chia-shu’s erosion strategy.
The Marine Corps has given up tanks and self-propelled artillery for drones and anti-tank gear, shifting toward defending the capital. The navy created a Coastal Operations Command, and the Coast Guard appointed a Navy vice admiral as deputy director-general in late September. Still, the authors flagged weak spots: the Joint Operations Training Base usually offers only battalion-level training, and limited airspace makes it hard for the Navy, Army Aviation and Air Force to train together. Taiwan has far fewer resources than China, they wrote, making joint operations essential.
Why This Matters:
Three stories, one thread: institutions decide whether power is checked or abused. Botvinick’s warning is that AI gives governments new ways to watch and control people, and that guardrails need building before abuses arrive, not after. The Mecca pact shows how security promises can outrun the hard work of governance, with no public text, no funding formula and no clear command, while energy disruptions at Yanbu and Hormuz reach ordinary consumers far from the Gulf. Taiwan’s debate shows a democracy preparing to defend itself while admitting its coordination gaps. In each case, transparency and accountability aren’t extras. They’re what keeps security from coming at the expense of the people it’s meant to protect.