Singapore and Sweden signed an AI cooperation plan. A declaration needs working projects

THE SIGNAL IN ONE SENTENCE

Singapore and Sweden have agreed to cooperate on trusted and innovative technologies. The announcement is broad enough to sound impressive. It is also broad enough to disappear into a filing cabinet unless somebody turns it into actual projects. During Singapore President Tharman Shanmugaratnam's state visit to Sweden, the two governments signed a Joint Declaration on Strategic Cooperation on Trusted and Innovative Technologies. Singapore's Ministry of Foreign Affairs published the declaration with its October 20 account of the visit. Sweden's government separately described the visit as an effort to deepen work on innovation, investment, sustainable energy and resilient infrastructure. The declaration names a formidable menu. It includes semiconductors and strategic hardware, Internet of Things technology, AI models and applications, quantum technology, digital and physical infrastructure, 5G and 6G connectivity, cyber security and post-quantum cryptography. For AI specifically, the two countries say exchanges may cover governance, public-sector digital transformation, regulatory approaches, workforce preparedness and practical measures for responsible use. They also plan to deepen engagement in standards-setting bodies and other international forums. That is the verified agreement. Here is what it is not. It is not a treaty. The declaration explicitly says it creates no rights or obligations under national or international law. It includes no financial commitment. It does not name a first AI project, assign a budget, announce a procurement, define a benchmark or report a deployment result. There is nothing sneaky about that. Declarations are often the diplomatic scaffolding that allows agencies, universities and companies to find one another. The problem begins when the scaffolding is photographed, praised and mistaken for the building. The plain signal is that every cooperation declaration needs a public project ledger. The ledger should begin with one very boring question: what is the work? If the countries want to compare AI governance, name the agencies doing the comparison, the laws and administrative practices being studied, the questions they will answer and the date of the first public output. If they want to improve public-sector digital services, identify one service in each country, its current problem, the people affected and the decision that will remain with a human official. If they want a workforce program, say who will be trained, by whom, for what task and how learning will be measured. Attendance is not an outcome. A hundred people in a room receiving a slide deck is international cooperation in the same sense that standing near a treadmill is cardio. The declaration's range makes prioritization especially important. AI models, chips, quantum systems, cyber security, connected devices and next-generation networks each have different institutions, supply chains, risks and time horizons. Folding all of them into one umbrella can help leaders see the connections. It can also produce a program where everyone attends the launch and nobody owns delivery. Give every project an accountable pair. One lead institution in Singapore, one in Sweden. Name the responsible program officials, not merely the ministers who welcomed the agreement. Record the universities, companies, unions, public agencies and community organizations involved. When ownership changes, update the record rather than letting the project become an orphan with a logo. Then publish the resources. The declaration itself commits no money, so future projects should distinguish existing staff time, new public funding, private contributions, shared facilities and in-kind support. A research exchange that uses an already funded laboratory is different from a joint grant program. A pilot built with donated cloud credits is different from durable public infrastructure. This matters because the cooperation areas are not cheap hobbies. Semiconductor research needs specialized equipment and supply-chain access. Quantum projects need hardware, technical talent and careful definitions of what is actually ready. AI evaluation can require secure data, computing capacity and domain experts. Trusted connectivity depends on physical networks, standards and long maintenance cycles. A ledger should show what each partner contributes and what happens when introductory support ends. If a pilot depends on temporary credits, a proprietary interface or one supplier's staff, write that down. The cleanest demo in Stockholm or Singapore can become an expensive souvenir when the free layer disappears. Public interest needs its own column. The declaration repeatedly invokes trust, resilience and responsible deployment. Those words should become tests rather than decorative adjectives. For a public-service AI project, record the service goal, legal authority, data used, affected population, accessibility requirements, language coverage, security review, human oversight and appeal path. Publish the baseline before the pilot starts. If the project is supposed to reduce processing time, measure current and future processing time. If it is supposed to improve access, show who was previously excluded and whether the gap actually narrowed. For governance exchanges, the output should be more than a conference summary. Compare how the two systems handle procurement, impact assessment, records, public explanation, vendor access, incident reporting and correction. Note where national rules differ. Cooperation does not require pretending that every regulatory choice is identical. The two countries can be useful mirrors for one another precisely because their contexts differ. Singapore is a city-state with highly centralized public administration and a strong role in Southeast Asian trade and digital policy. Sweden operates inside the European Union's regulatory system and has its own industrial, research and public-sector institutions. A practice that transfers well should survive those differences. A practice that does not transfer should be documented rather than politely declared universal. That is where small, paired projects beat giant slogans. Choose one bounded problem. Test the same reporting template in both countries. Compare the administrative burden, technical result and public response. Publish what failed. Then decide whether the work deserves expansion. Standards work needs similar specificity. The declaration mentions cooperation in multilateral, regional and standards-setting bodies. A project ledger should identify the forum, proposal, responsible delegation and public position. It should distinguish a technical contribution from a political statement and disclose relevant commercial interests. That does not mean publishing confidential negotiating instructions. It means letting the public see whether a promise to shape trusted technology produces an actual submission, reference implementation, test method or governance proposal. The ledger also needs a stop column. Diplomatic programs tend to accumulate. A workshop leads to a working group, the working group leads to another meeting and the absence of a result becomes a reason to keep meeting. Every project should carry a next decision date and a rule for continuing, changing or closing the work. Useful closure is not failure. If a joint team learns that a shared data system creates unacceptable privacy risk, that finding can save both governments from a bad procurement. If an AI pilot produces no improvement over a simpler process, publishing that result is better cooperation than hiding it behind a celebratory paragraph. Finally, keep the declaration visible beside the ledger. Readers should be able to move from the original intention to the funded project, evidence, decision and outcome. The record should show which parts of the declaration have active work, which remain exploratory and which were never pursued. Singapore and Sweden have given themselves a useful field of possibilities. Their complementary strengths in public administration, industrial technology, research and international standards could produce projects with value beyond either country. But possibility is not delivery. The difference is a named problem, a responsible owner, a real budget, a public-interest test and evidence anyone can inspect. Put those five things in a ledger and a diplomatic declaration can become working infrastructure. Leave them out and the plan remains exactly what it is today: a promising list with very good stationery.