(Joshua Sukoff/Shutterstock)
Future historians will ask why the world spent so much time worrying about the profound impacts of AI yet did so little to govern or guide it. Part of the answer lies in deliberate strategies of distraction that have worked well, at least until now, in manipulating politics and the media.
These strategies – visible again this summer in new initiatives from Mark Zuckerberg, Demis Hassabis and Elon Musk – were answers to a contradiction that every leading figure in AI knows well.
To raise capital, figures such as Zuckerberg, Hassabis and Musk, if not the whole of Silicon Valley, have to emphasise the transformative power of AI. But if the technology is even a fraction as transformative as they claim, then it’s hardly plausible that private companies or tech oligarchs will be best-placed to guide it. So what to do?
Their answers to this obvious contradiction have come in a playbook with seven main moves.
1. Distract: The first move dominant tech players make is to pursue low-impact, but interesting initiatives. For instance, setting up ethics centres in universities (but never centres on governance) or focusing on the safety of foundational models, nearly always as an alternative to government legislation rather than as a complement to it. Creating distracting ethics bodies has the effect of pulling in many of the brightest minds, which might otherwise be working on more serious, enforceable rules.
Rishi Sunak’s AI summit at Bletchley in 2023 was a prime example, launching an AI Safety (later Security) Institute which tests foundational models. This is important work, but addresses only a tiny fraction of the risks associated with AI. Several years later the initiatives have had little real-world impact (the first warnings about Anthropic’s Fable and Mythos models, for example, came from Amazon, not from one of these institutes).
2. Amplify: Next, Big Tech players talk a lot about existential risk, issuing vague statements every few years. These statements fill the media with content about the possible end of humanity, while diverting journalists and politicians from thinking about present-day governance, including facial recognition, targeted advertising and copyright abuses.
A gathering of AI pioneers in Puerto Rico in 2015 set the tone. Figures including Hassabis, Stuart Russell, Max Tegmark, Jaan Tallinn and Musk came together to discuss what rules might constrain AI. Their open letter committed to preparing for the risks that it could bring: “because of the great potential of AI, it is important to research how to reap its benefits while avoiding potential pitfalls … [through] expanded research aimed at ensuring that increasingly capable AI systems are robust and beneficial: our AI systems must do what we want them to do.” But nothing was done to act on these sentiments. A succession of equally empty statements has followed every few years, with equally little real-world impact.
3. Mystify: Third, those with power have emphasised the magical aspects of AI. Sam Altman of OpenAI describes artificial general intelligence as a “magic intelligence in the sky,” while Hassabis describes it as a tool to “solve everything.” Musk promises that it will deliver super-abundance (so that it won’t matter that there aren’t any jobs left). Karl Popper, the Austrian philosopher, argued that we should “work for the elimination of concrete evils rather than for the realization of abstract goods”. The AI distraction playbook does the opposite.
4. Feint: Many tech leaders talk publicly about the need for regulation, while lobbying in private against it. They give every appearance of being open to dialogue and debate, floating interesting and radical new policies on AI. For example, many tech figures suggest providing a universal basic income or allocating AI shares to the wider population, confident that there is little prospect of these things happening, while never conceding an inch in relation to real regulation.
5. Co-opt: Many tech companies try to get their people in key places serving as lobbyists in Washington, Brussels and London. Powerful players ensure they work hard to keep out challenging voices or experts. The capture of the now-closed Department for Science, Innovation and Technology (DSIT) in the UK is a fascinating case study.
6. Threaten: Many companies use subtle, or not-so-subtle, threats. Warn any country that stands up to you that investment will be removed and it will be left behind in a vital technology for future growth (a threat which is not implausible).
7. Block: Finally, and perhaps most importantly, many tech players are doing all they can to block the creation of any institutions that could potentially challenge them or offset their power. Silicon Valley now favours committees, conferences, summits, and high-level expert panels – but at all costs maintains the asymmetry between strong, wealthy, highly organised corporate power and weak and poorly organised public power.
The technique is now to present anyone arguing differently as a lover of bureaucracy, as anachronistic and ignorant about technology. They argue for self-regulation, as Hassabis recently did, though there are no examples where self-regulation has proven adequate for powerful technologies (there is a good case for mobilising the big companies to check and test each other’s models for flaws – but no case for doing so without any kind of public backstop).
(TechCrunch/Wikimedia Commons)
These strategies – to distract, amplify, mystify, feint, co-opt, threaten and block – succeeded in much of the world, but not everywhere. China introduced ever more elaborate laws throughout this decade (undercutting the argument that you can either have frontier technologies or regulation, but not both). For instance, it has some of the toughest laws on algorithmic and software vulnerability disclosure while also pioneering the ultra-cost-efficient LLM DeepSeek. In the EU, the regulatory juggernaut eventually produced the AI Act, which started being implemented in 2024, though there is still only one institution to regulate AI (the Spanish Agency for the Supervision of Artificial Intelligence).
But overall, the strategies remain in place, eagerly pursued by large teams of lobbyists. Around 3,500 lobbyists work on AI in Washington alone. Alphabet and Meta each employ nearly 90 federal lobbyists on the Hill and the industry’s federal lobbying budget is estimated at over $40m in the first half of 2026 alone. Huge sums are now going into Super PACs to influence elections. Hundreds of commentators have effectively been bought off while philanthropy has been largely silent or complicit in the distraction strategies.
There are many well-intentioned initiatives around AI governance underway. What’s striking is how few seriously challenge the above playbook. Some new bodies are heavily influenced by liberal and libertarian thinking, arguing that any role for government would contravene individual freedom. As with the Internet, in the 2000s and 2010s, this logic clears the way for massive autocratic corporations to accumulate power.
Regardless of how AI impacts the world, both for good and bad, the success of these strategies will go down in history, alongside those of the tobacco industry and Big Oil, which also needed to confuse and distract (the cases are very different but the echoes are striking).
However, the success of the playbook may at last be fading, mainly because the politics of AI is now changing. This is evident, particularly in the US, where AI is rapidly becoming far more contested (50% of Americans are more concerned than excited about AI, compared to 10% who say the opposite). Sovereignty – and the need for sovereign AI – is gaining far more attention, especially after the temporary withdrawal of Anthropic’s frontier models back in June. Additionally, many of the everyday risks – from misinformation to cognitive decay – that the distraction playbook tried to ignore, are gaining attention from the general public.
But the extraordinary asymmetry between corporate power and public and civil society is not about to disappear. I hope we will see more politicians achieve a better balance of AI enthusiasm and scepticism, focused on how to use the many powers they do have – over law, regulation, procurement and public education. Just this week, William Hague commented that “how we use and control AI is now becoming the single most important issue in the world”. He’s right, but far too many politicians have made similar comments in the past and not pursued the serious work needed to shape AI. In effect, they have defaulted to being followers, not leaders, creating a vacuum that the tech companies are only too willing to fill.
