AI tools are just a handful of the countless ways that our lives are already governed by digital platforms and the companies that control them, writes Wendy H. Wong.Manuel Ausloos/Reuters
Wendy H. Wong is a professor of political science and Principal’s Research Chair at the University of British Columbia’s Okanagan campus, and the author of We, the Data: Human Rights in the Digital Age, which won the 2024 Balsillie Prize for Public Policy. Her next book, Mirrored States, will be published in 2027.
It seems AI developers can’t stop making potentially dangerous advancements in their fields and then sounding the alarm.
The latest warning came last week, when a former Anthropic and OpenAI employee named Jacob Coxon wrote on X that “the people building AI earnestly believe that it could kill us all by the end of the decade.” (One of Mr. Coxon’s former colleagues at Anthropic, Evan Hubinger, further fanned the flames when he wrote “Jacob is correct here–we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade.” This came after more than 1,000 AI scientists signed a letter in July that urged policy makers to “pace” AI.
OpenAI and Anthropic have been leaders in a breathless race for AI dominance for the past few years – a race against one another, against Chinese companies, and against other U.S.-based rivals. Will they be able and willing to “pace” AI?
Anthropic CEO urges AI companies to slow development as safety concerns mount
Only if we see the AI problem for what it is: the inability to separate physical life from digital platforms. This lack of separation creates the vulnerabilities identified in another recent open letter, “A call for collective action on cyber defense,” signed by more than 100 companies: “The companies and public services our communities depend on – from hospitals to water treatment plants to the infrastructure that powers the internet – are at risk” because so much of what we do now is enabled by and dependent on digital platforms.
What we need now is to reframe the AI problem as a “digital platform” problem. Our dependence on digital tools makes us particularly vulnerable to AI.
Our lives have been and increasingly are dependent on Big Tech companies’ many and varied platforms. AI tools are just a handful of the countless ways that our lives are already governed by digital platforms and the companies that control them.
Apps are our means of connectivity to the world. The companies that make and distribute them also control the means of computation that has become so valuable to the world: the data, the algorithms and the computing resources necessary to make large language models (LLMs) spit out the words or pictures we want, or to leverage agentic AI to optimize workflows.
Think about your day and how many platforms you need or choose to use. We confess our sexual attraction, book vacations, share our musical tastes, order school lunches, manage bank accounts and insurance coverage, research current events and hobbies and check medical results online.
We are in trouble, but not because of the possibility that AI is going to take over and start killing humanity – which, to be clear, is something not all AI scientists agree will happen any time soon.
Anthropic CEO Dario Amodei at the AI Impact Summit, in New Delhi, India, in February. In an essay published to social media last weekend, Mr. Amodei outlined a three-pronged plan for AI pacing.Bhawika Chhabra/Reuters
In an essay published last weekend, Anthropic CEO Dario Amodei wrote he worries botnets can take over the entire internet and become more powerful and destructive still than what we’ve already seen. The reason to be scared isn’t because AI bots will develop unfettered and become some version of Skynet from The Terminator. The reason to be worried is that our lives are so integrally tied into digital platforms – and these platforms are inextricable from our physical existences.
In his essay, Mr. Amodei outlines a three-pronged plan for AI pacing: embedding third-party evaluation, and two parallel kinds of co-ordination: between AI companies to decide how to pace AI, and among global policy makers.
AI’s biggest players are self-serving. But we must heed them anyway and pause development
I appreciate that Mr. Amodei calls for co-operation among companies and among governments. But separating these two processes is untenable. AI, and by extension any technology that is connected to the internet, is not something we can hive off from “regular” policy making and human life. They are integrally connected, from agriculture to banking to communications systems.
Frontier AI companies cannot be left to their own devices in a separate corporate process of unelected policy making. Self-policing rarely works. We have a better shot at regulation if industry-sourced standards involve policy makers and rest on the enforcement capacities of the state. My colleague David Lake and I have written about professionalization as a way for AI developers to set its own standards but have external enforcement (akin to doctors, lawyers and many other professions).
Governments ostensibly work in the public interest. They cannot determine the public interest without knowing what AI companies can do, want to do and how those capabilities and hopes might harm people.
They also cannot decide the public interest by relying on the market leading makers of AI to tell them the truth. Other companies need to be there, yes, but also experts whose interests aren’t directly tied to shareholders: technologists, but also civil society, social scientists and the arts.
Serving the public interest also means hearing from those who are opposed to AI, for the many reasons that people have grown wary of the technologies. Will AI “cure” cancer or solve climate change because it’s really good at solving math problems? Last I checked, cancer and climate change are complex problems with many, many tricky choices and hinge on a host of known and unknown factors that shape possible trajectories.
Opinion: It’s time Ottawa faced up to AI’s catastrophic risks
Big Tech companies are really good at making digital platforms to provide connectivity. They are really good at improving computational capacity out of tiny silicon-based chips. Big Tech companies are not good at working in the public interest.
The public interest has been lost in these conversations as policy makers chase what AI and other Big Tech companies may give. Instead, governments should be conditioning – through governance – what the potential harms and benefits of Big Tech companies might produce, based on the public interest.
Much attention has been focused on the “sticks” of governance: liability lawsuits, fines, antitrust cases. These are time and resource intensive tasks with uncertain results. We need more “carrots” of regulation. Governments need to bring AI companies to the table to create enforceable policies that enable the benefits of AI without conceding to the wildest prophecies of the technology’s makers.
One underutilized carrot is procurement – when the government goes shopping. Officials need to start being more demanding, discriminating consumers of digital technologies, including AI. Contracts need to be conditioned in the public interest and democratic values. Should governments purchase automated license plate readers? Only if the data can be protected (which is questionable with Flock cameras) and the risk of data leakage or algorithmic abuses are outweighed by social benefits.
When there is an app for everything and we have all come to expect to have answers instantly, the problem when AI systems and companies “go rogue” isn’t just an AI problem.
It’s a deeply human problem.
Our difficulty is that our digital-analogue lives cannot be separated. Therefore, who decides the future of AI governance, should not be left to the few who have developed the tech that got us here. Governments are the public interest custodians that can govern Big Tech companies. They need to step up their role.