Should we try to patent the sun?

We rarely speak about our right to science. In the age of AI for Science, the use of AI to accelerate, augment, or transform scientific discovery, we should. As we implement our new AI for All national AI strategy, this right is under pressure.

In 1948, the Canadian drafter of the Universal Declaration of Human Rights, John Humphrey, included Article 27: everyone, it says, has the right to freely participate and “to share in scientific advancement and its benefits.”[1] It was a radical idea then, and it remains one now.

The fruits of discovery and research, the drafters insisted, belong to all of us, not only to those who own the laboratories. Eighteen years later, the world made that promise binding through the International Covenant on Economic, Social and Cultural Rights (ICESCR), which 169 countries, including Canada, have since joined.[2]

More recently, AlphaFold2, the Google DeepMind model that predicted the 3D structures of more than 200 million proteins and earned its creators a share of the 2024 Nobel Prize in Chemistry,[3] was built on computing power few researchers can access. Training it required 128 TPUv3 cores for 11 days, on proprietary hardware available only through Google’s own cloud. Generating these predictions also demands heavy resources, putting such work out of reach for many labs.

If even well-funded labs struggle to reach this frontier, ordinary citizens stand further still from it. Most Canadians are unlikely to operate, inspect, or benefit from these advances, in part because AI literacy and public trust remain thin. Pillar 2 of the National Strategy promotes AI literacy, but it casts Canadians as users and adopters rather than as participants in governance and development. That is the paradox of scientific progress in our time. AI has never been more promising, or more narrowly held.

History shows what science is worth sharing

When Jonas Salk was asked in 1955 who owned the patent on his polio vaccine, he answered, “Well, the people, I would say. Could you patent the sun?”[4] The vaccine was never patented, and polio has since been driven from all but a handful of places.

When Human Genome Project scientists adopted the Bermuda Principles in 1996, releasing every fragment of sequenced DNA into the public domain within a day, they chose to treat the code of life as a common good. Economists later estimated that genes locked behind private intellectual property generated up to a third less follow-on research.[5]

The reverse is equally well documented. For years, life-saving HIV medicines existed but stayed out of reach of the people dying without them, until a global access campaign and the flexibilities affirmed in the 2001 Doha Declaration on TRIPS and Public Health forced generic prices down.

Two decades later, the pattern repeated. During COVID-19, wealthy nations bought up most of the vaccine supply while much of the world waited, a division the World Health Organization’s own officials called “vaccine apartheid.”[6] The science and know-how existed. What failed was access and benefit-sharing.

This is precisely why the right to enjoy the benefits of scientific progress cannot be a courtesy extended only in good times: it is needed most as geoeconomic, planetary, and human security pressures mount.

A fragmenting world of science, technology, and innovation is closing the doors

This should worry anyone who takes the right to science seriously: the world is fragmenting just as science grows more consequential. The United Nations committee that interprets the ICESCR warned, in a 2020 statement, that “fragmented national responses to these transnational technologies” would “perpetuate technological divides and economic disparities.” It named AI directly.[7] Since then, the warning has sharpened into fact.

The collaboration that drives modern AI science is thinning. Research ties between the United States and China, the world’s two largest producers of knowledge, are fraying under export controls, security investigations, and mutual suspicion. Scientists have been constrained and partnerships severed. After Russia’s invasion of Ukraine, even CERN and the International Space Station, symbols of research across national divides, severed cooperation with Russian institutions.[8] Many of these barriers stem from real or perceived threats, yet they also narrow the freedom to participate in and benefit from research.

The same instruments that block a dangerous transfer may also wall off the ordinary, cumulative work across borders on which much of discovery depends.

Nowhere is the new divide starker than in AI itself. Africa and Latin America together hold roughly three percent of the world’s AI computing capacity, despite being home to roughly a third of the world’s people.[9] Some 2.6 billion people, a third of humanity, were still offline in 2023.[10] The instruments of AI discovery are being built, owned, priced, and governed by a handful of firms in a handful of countries, while the rest of the world may use them but not own them.

Our choices ahead: the right to science shows the way through

To speak of science as a right is to say that rights carry obligations. The right to science does not mean that every dangerous dual-use technology is shared with everyone. It calls for discipline in making access the default, with justified exceptions, so that progress reaches beyond the strong few. For Canada and other middle powers, building alliances with like-minded and non-traditional partners is a pragmatic way forward. It gives us a standard against which to measure technoscientific nationalism, vaccine stockpiling, and the rapid enclosure of AI for Science.

There are signs of movement in this direction. Global dialogues on AI governance are considering financing to widen access to computing power, data, and talent. India has built a public pool of thousands of processors and opened it to its own researchers and startups at subsidized cost.[11] Canada and the United Kingdom, with African partner universities, support the Africa Compute Initiative. These are early steps, but they point the right way in treating the foundations of AI as global public infrastructure, much as an earlier generation treated the polio vaccine and the human genome.

Even so, a tension runs through Canada’s recently released National AI Strategy. Pillar 6 emphasizes “building trusted partnerships and global alliances,” recognizing that governance and development require global cooperation. However, the broader strategy is framed through digital sovereignty, most visibly in Pillar 4, which builds a “sovereign AI foundation” where compute, cloud, connectivity, data, and talent must remain under “Canadian control, on Canadian terms.” Rather than a shared endeavor whose benefits must be diffused, AI is conceived as a strategic resource whose value lies in national competitiveness, in “scaling Canadian champions” and sovereign capabilities.

The choice before us is not new, but it is urgent. Every generation decides whether its greatest scientific instruments will be a shared inheritance or a private possession. Salk chose not to patent the polio vaccine. Genome scientists chose the commons.

Canadians face the same choice with AI in a more fragmented world, and we need to open this conversation. The right to science, written by a Canadian, is the most fundamental guide we have. It holds that everyone should enjoy the benefits of discovery while the moral and material rights of inventors are protected. The only question left is whether we still intend to uphold our 1948 commitment to ensuring that AI’s discoveries benefit everyone in making AI for all.

References

[1] United Nations, Universal Declaration of Human Rights (1948), Article 27. un.org. The first draft was prepared by the Canadian legal scholar John Humphrey.

[2] International Covenant on Economic, Social and Cultural Rights (1966), Article 15(1)(b). ohchr.org. Confirm the current number of states parties (reported as 169) against the UN treaty-body database before publication.

[3] The Royal Swedish Academy of Sciences, “The Nobel Prize in Chemistry 2024” (Hassabis and Jumper for AlphaFold; Baker for protein design). nobelprize.org.

[4] Jonas Salk, interview with Edward R. Murrow, “See It Now,” 12 April 1955. Widely reported and genuine; note that the vaccine’s lawyers had separately concluded it likely could not be patented.

[5] Heidi L. Williams, “Intellectual Property Rights and Innovation: Evidence from the Human Genome,” NBER Working Paper 16213 (2010), finding roughly 20 to 30 percent reductions in follow-on research and product development for genes held under Celera’s IP. nber.org.

[6] On COVID-19 vaccine inequity and the TRIPS-waiver debate (India and South Africa proposal, Oct. 2020; WTO compromise, June 2022), and WHO officials’ use of “vaccine apartheid.”

[7] UN Committee on Economic, Social and Cultural Rights, General Comment No. 25 (2020) on science and economic, social and cultural rights (Article 15), UN Doc. E/C.12/GC/25, paras. 72 to 74. digitallibrary.un.org.

[8] CERN Council decision to let cooperation agreements with Russia and Belarus lapse (Russia agreement ended 30 November 2024).

[9] Brookings, on AI-compute concentration: Africa and Latin America together account for about 3% of global AI compute capacity. brookings.edu.

[10] International Telecommunication Union, Facts and Figures 2023 and 2024: an estimated 2.6 billion people (about 33% of the global population) offline in 2023. itu.int.

[11] India’s IndiaAI Mission: a subsidized public pool of graphics processors (reported at roughly 18,000 GPUs) for domestic researchers, startups, and academics.

References

[1] United Nations, Universal Declaration of Human Rights (1948), Article 27. un.org. The first draft was prepared by the Canadian legal scholar John Humphrey.

[2] International Covenant on Economic, Social and Cultural Rights (1966), Article 15(1)(b). ohchr.org. Confirm the current number of states parties (reported as 169) against the UN treaty-body database before publication.

[3] The Royal Swedish Academy of Sciences, “The Nobel Prize in Chemistry 2024” (Hassabis and Jumper for AlphaFold; Baker for protein design). nobelprize.org.

[4] Jonas Salk, interview with Edward R. Murrow, “See It Now,” 12 April 1955. Widely reported and genuine; note that the vaccine’s lawyers had separately concluded it likely could not be patented.

[5] Heidi L. Williams, “Intellectual Property Rights and Innovation: Evidence from the Human Genome,” NBER Working Paper 16213 (2010), finding roughly 20 to 30 percent reductions in follow-on research and product development for genes held under Celera’s IP. nber.org.

[6] On COVID-19 vaccine inequity and the TRIPS-waiver debate (India and South Africa proposal, Oct. 2020; WTO compromise, June 2022), and WHO officials’ use of “vaccine apartheid.”

[7] UN Committee on Economic, Social and Cultural Rights, General Comment No. 25 (2020) on science and economic, social and cultural rights (Article 15), UN Doc. E/C.12/GC/25, paras. 72 to 74. digitallibrary.un.org.

[8] CERN Council decision to let cooperation agreements with Russia and Belarus lapse (Russia agreement ended 30 November 2024).

[9] Brookings, on AI-compute concentration: Africa and Latin America together account for about 3% of global AI compute capacity. brookings.edu.

[10] International Telecommunication Union, Facts and Figures 2023 and 2024: an estimated 2.6 billion people (about 33% of the global population) offline in 2023. itu.int.

[11] India’s IndiaAI Mission: a subsidized public pool of graphics processors (reported at roughly 18,000 GPUs) for domestic researchers, startups, and academics.