“When I look at AI, I see only monopolies. Nothing else.” These extraordinary words from Andreas Mundt, head of the German Competition Agency, capture a mounting European push to regulate the artificial intelligence industry.

Our Competition and Markets Authority (CMA) exemplifies the shifting tides. Last September, it published an upbeat report about the artificial intelligence foundation model market, which includes ChatGPT, Snapchat My AI and Bing Chat. Praising the sector’s “vibrant competition,” it spotlighted how 160 foundational AI models had been created by giant tech companies like Google and Meta, alongside new firms like OpenAI, Anthropic, and Midjourney.

This was a highly dynamic industry, the CMA concluded, where models “can quickly become obsolete.” The market was also diverse and pluralistic, producing both open source and proprietary models under varied ownership structures, including vertically integrated firms, downstream partnerships, payments for model access, or third-party plug-ins.

The CMA was thus bullish on AI’s potential to revolutionise downstream products like search, healthcare, and energy, while creating entirely new markets too. Although cognisant that dominant firms might one day exercise anticompetitive power over key inputs like data, it avoided assuming this outcome. Intervention should be “proportionate and targeted at identified risks,” it said, as harms “begin to emerge.”

Since then, however, the CMA has become worryingly fatalistic about AI competition. Its recent “update” bemoaned the “growing presence” of Google, Amazon, Microsoft, Meta, Apple, and Nvidia (“GAMMAN”) within the AI “value chain.” Big Tech’s involvement is seen as inherently problematic to fair competition. The regulator yesterday added to its investigation of Microsoft’s partnership with OpenAI by inviting comment on whether Amazon and Anthropic’s partnership, Microsoft’s (tiny) investment in Mistral AI and Microsoft’s mere hiring of Inflection AI’s employees breached UK merger rules.

Amazon called this action “unprecedented”. The shift is certainly jarring because competition clearly hasn’t weakened since September. Indeed, the CMA’s update acknowledges 120 extra AI models since then, including Anthropic’s Claude 3 and Google’s Gemini, alongside numerous open-source products. So why, exactly, are six major tech companies investing billions to advance the frontiers of AI within a highly competitive sector being treated like harmful monopolies?

Perhaps in atoning for missing Big Tech’s dominance in existing digital markets, these firms occupy psychological monopoly status at the CMA. It certainly obsesses with how they “could” or “may” harm AI competition, hence yesterday’s actions that extend to worrying about arms-length partnerships or even firms hiring personnel. At times, however, this knee-jerk opposition gets perilously to analogising that it would be worrisome for petrol car producers like VW or Mercedes to compete with Tesla in developing electric vehicles.

Indeed, the CMA’s recent anti-Big Tech hostility is striking. It fears, rather than champions, the consumer benefits to integrating AI into existing tech ecosystems. It downplays how mergers or partnerships with giant firms can help scale or incentivise innovative AI start-ups too. Rather than addressing anticompetitive acts when they arise, it seems the CMA wants to design the sector’s structure to meet its own conception of healthy competition.

This is highly premature given the industry’s rapid evolution (mere months ago, everyone thought OpenAI and Google would dominate). Determinism in a nascent market is a dangerous way to regulate: Facebook’s dominance in social media post-2010 didn’t justify regulating MySpace in 2003, for example. So I fear a heavy-handed CMA will sully the UK’s reputation as an open environment for this maturing industry, undercutting the success of Deepmind and Microsoft locating here and our fruitful AI start-up scene.

Unfortunately, this is what happens when regulators obsess about “Big Tech” firms rather than keeping laser focused on market outcomes for customers. As one former US Federal Trade Commissioner put it, we’re hurtling towards a “Yogi Berra” competition policy. “Nobody ever goes there, it’s too crowded,” remarked Berra. On AI, the CMA and Mundt fret about “all those Big Tech monopolies, competing.”