Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Eric Schmidt’s warning is partly right: the EU’s AI Act adds real compliance work and can slow launches, especially for foundation-model providers and companies in high-risk sectors. But regulation is not the sole reason Europe struggles to produce frontier AI companies. Britain’s opening is genuine only if it pairs flexible rules with compute, energy, capital, talent, procurement and access to European customers.
What Eric Schmidt actually argued
Schmidt made his comments in a BBC Radio 4 Today interview on 13 February 2025, shortly after the Paris AI Action Summit. He praised European and British researchers but argued that Europe lacks sufficiently integrated markets, very large technology companies and a permissive enough environment for private investment. His fear was that the AI revolution would be invented elsewhere.
He pointed to DeepMind’s origins in London as evidence that Britain can produce exceptional technical talent. He described a closer US–UK relationship as a potentially “winning strategy” and said Brussels was holding Europe back with strong regulation. Those are Schmidt’s judgements, not an established finding or a formal UK–US policy alliance. (ITPro report)
What the EU AI Act requires
The AI Act is not a blanket ban on artificial intelligence. It is a risk-based law: prohibited practices face the toughest restrictions, while obligations become more extensive as the potential harm rises. A low-risk productivity tool is not treated like an AI system used in employment, healthcare, education, critical infrastructure or public services. (Council of the EU overview)
#1 Best Overall
The timetable
| Date | What applies |
|---|---|
| 1 August 2024 | The Act entered into force. |
| 2 February 2025 | Prohibited AI practices and AI-literacy obligations began applying. |
| 2 August 2025 | Obligations for providers of general-purpose AI models began applying. |
| 27 July 2026 | The AI Omnibus entered into force, simplifying parts of the regime and changing transition arrangements. |
| 2 August 2026 | Most remaining rules and enforcement began applying, subject to exceptions and revised deadlines. |
| 2 December 2026 | Some transparency-related deadlines are scheduled to move under the revised timetable. |
| 2 December 2027 | Stand-alone high-risk AI rules and national AI regulatory sandboxes are scheduled to apply. |
| 2 August 2028 | High-risk AI embedded in regulated products is scheduled to apply. |
The exact transition applicable to a product depends on its category, provider status and sector. The European Commission’s implementation timeline should be checked for a specific deployment.
Where the work falls
- Foundation-model providers: general-purpose AI obligations can include technical documentation, information for downstream users, copyright policies and, for the most capable models, systemic-risk duties.
- High-risk providers and deployers: requirements can include risk management, data governance, testing, human oversight, logging, accuracy, cybersecurity and conformity assessment.
- Transparency obligations: certain systems must disclose that people are interacting with AI or that content has been artificially generated or manipulated.
- Prohibited uses: some practices are banned outright, rather than managed through paperwork.
- Enforcement: the framework combines EU-level and national responsibilities, with penalties and compliance expectations varying by breach and actor.
A small company deploying a third-party model for an ordinary internal workflow therefore faces a different exposure from a model developer or a medical-device manufacturer. Calling the Act simply “red tape” hides those distinctions.
Could regulation stifle innovation?
The strongest case against the Act
Critics have a plausible mechanism. Compliance can consume scarce engineering, legal and testing capacity. Unclear guidance or delayed technical standards can make a launch date uncertain. A startup may postpone a feature, offer a less capable version in Europe or maintain separate processes for different markets. High-risk conformity assessments are particularly difficult when the underlying models and standards are changing quickly.
Those are risks, not proven economy-wide outcomes. There is not enough evidence to say that the AI Act has already caused companies to leave Europe or that it has measurably reduced innovation. Such claims require company-level launch, investment or product data.
The case against blaming Brussels alone
Europe’s scale-up problem predates the Act. Schmidt himself separated fragmented markets and the difficulty of building very large technology companies from regulation. Limited late-stage capital, expensive energy, inadequate compute, cautious procurement and weak commercialisation of research can constrain a company before legal compliance becomes its main obstacle.
Nor is the EU pursuing regulation in isolation. Its wider AI policy includes AI factories, adoption programmes, sandboxes and an AI Continent strategy. The AI Omnibus, which entered into force on 27 July 2026, extends some deadlines, broadens relief for smaller firms and expands experimentation provisions. (European Commission announcement)
Rules can also create value. A common framework may give investors and enterprise buyers clearer expectations, reduce the need to navigate 27 entirely different regimes and increase trust in high-impact deployments. The UK’s own action plan says clear rules and credible assurance can support investment and adoption, while ineffective regulation can hold sectors back. (UK AI Opportunities Action Plan)
Recommended Free Tools
How the UK and EU approaches differ
| Issue | EU | UK |
|---|---|---|
| Overall model | Cross-sector, risk-based statutory framework through the AI Act. | Opportunity-led strategy using existing regulators, targeted rules and assurance arrangements. |
| General-purpose models | Dedicated obligations began applying on 2 August 2025. | A more consultative and flexible approach, with future arrangements still developing. |
| High-risk systems | Detailed legal duties, with revised transition dates. | Sector regulators remain central to implementation. |
| Innovation support | AI Act sandboxes plus EU capability and adoption programmes. | AI Growth Zones, sovereign compute plans and public-sector deployment. |
| Market position | Large single market with complex implementation. | Smaller domestic market, close US links and continued exposure to EU rules. |
| Principal risk | Compliance complexity and slow guidance. | Insufficient certainty, scale, infrastructure or safeguards. |
Britain is not an unregulated alternative. Its model is more flexible and decentralised, but it still depends on safety, assurance, sector regulation and potentially new legislation. (UK one-year progress report)
Why Britain has a real opening
Britain starts with assets that Schmidt highlighted. DeepMind was founded in the UK; Arm and Wayve are prominent home-grown technology companies; and the country has strong universities, financial markets and links to US investors. The government’s 2025 action plan described the UK as the world’s third-largest AI market and set out 50 recommendations covering infrastructure, skills, adoption and public services. That ranking is the government’s characterisation, not an independently established measure.
The January 2026 progress report says 38 of the 50 commitments had been completed after one year. It cites AI Growth Zones, expanded public-sector compute, a National Data Library, skills measures and a Sovereign AI Unit backed by up to £500 million. The government also set an original target of at least a 20-times expansion in AI Research Resource capacity by 2030. These are programmes and targets, not proof that Britain has already become a frontier-AI superpower.
Potential advantages
- English-language research and commercial networks that connect readily to US firms and capital.
- A concentrated national market where government can act as an early customer.
- Universities, specialist researchers and London’s finance, legal and venture ecosystem.
- Regulatory flexibility that may allow faster experimentation than a fully centralised rulebook.
- The possibility of serving as a bridge between US model companies and European users.
The price of regulatory divergence
A UK company does not escape EU obligations merely because it is incorporated in Britain. Selling an AI system into the EU, placing a covered model on that market or supplying a regulated product can trigger duties under the Act. A provider may therefore need one compliance architecture that satisfies EU requirements alongside UK and other international expectations.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
That creates a practical trade-off. Divergence can make British experimentation faster, but separate documentation, testing and product controls can raise costs for firms that need European customers. The EU’s 450-million-person market is also too important for most scale-ups to ignore. “Light touch” is an advantage only if it means faster, clearer decisions—not uncertainty that forces every company to seek private legal interpretations.
Best Value
What Britain must do to convert opportunity into results
- Build compute: expand public and commercial access to advanced accelerators, with transparent allocation for researchers and startups.
- Secure affordable energy: data-centre capacity is useless if power connections, prices and planning remain bottlenecks.
- Unlock scale-up capital: early grants cannot substitute for patient growth finance and credible routes to global listings or acquisition.
- Retain talent: immigration, research funding and attractive technical careers must keep frontier researchers in Britain.
- Create demanding customers: public procurement should reward safe, interoperable systems that solve measurable problems rather than fund pilots indefinitely.
- Clarify copyright, liability and data access: developers and users need rules that are predictable enough to price risk.
- Make sandboxes useful: testing environments should provide timely regulator feedback without implying exemption from eventual duties.
- Build credible assurance: independent evaluation can reassure buyers without imposing identical controls on every low-risk application.
- Stay interoperable with Europe: British firms need practical guidance on meeting EU requirements without maintaining wholly separate products.
- Commercialise research: universities need incentives, management expertise and capital to turn discoveries into durable companies.
Paris and the limits of a US–UK “winning strategy”
The political context matters. At the February 2025 Paris summit, the United States and United Kingdom did not sign the final international declaration. ITPro reported that the UK cited concerns about practical clarity, global governance and national security. That episode illustrates a divide between innovation-first rhetoric and more precautionary or rights-focused approaches, but it does not establish a formal US–UK regulatory alliance.
Britain also cannot match US companies’ capital, domestic demand or compute simply by aligning politically with Washington. Its most realistic advantages may be research, specialist applications, trusted deployment, public-sector adoption and brokerage between markets—not necessarily producing the next company on the scale of the largest US laboratories.
Verdict: Schmidt identifies a risk, not a complete diagnosis
Schmidt is right that Europe risks regulating from a position of insufficient technological power. The AI Act creates genuine obligations, and implementation quality will determine whether a common market becomes an asset or a delay. But lighter regulation alone would not solve Europe’s fragmented markets, weak scale-up finance, compute shortages, energy constraints or commercialisation gap.
Britain has a credible chance to become Europe’s fastest-moving major AI economy, a trusted bridge to US technology and a strong customer for advanced systems. That outcome is conditional. It requires the government’s programmes to become infrastructure, investment and deployed products while preserving enough regulatory clarity and EU interoperability to keep the market open.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
