If your business uses AI and serves customers anywhere in Europe, the EU AI Act is your problem too. Brexit did not build a firewall around your compliance obligations. The Act has extraterritorial reach, and it catches UK companies in exactly the same way GDPR did.

The EU AI Act works on a market-access basis, not a headquarters basis. If your AI system produces outputs that are used by people located in the EU, you fall within scope. That includes:

  • AI-powered customer service tools used by EU clients
  • Automated decision-making in HR, finance, or insurance affecting EU residents
  • AI models embedded in products sold into EU markets
  • Chatbots, recommendation engines, or scoring systems accessed from EU member states

The enforcement mechanism mirrors GDPR. You do not need an EU office to be liable. You just need EU-based users.

The Act creates four tiers of risk, and your obligations scale accordingly:

Unacceptable Risk (Banned)

These are already prohibited as of February 2025. Social scoring systems, real-time biometric identification in public spaces (with narrow law enforcement exceptions), and AI that manipulates behaviour through subliminal techniques. If you are doing any of this, stop.

High Risk (Heavy Obligations)

This is where most UK businesses need to pay attention. High-risk systems include:

  • AI used in recruitment, CV screening, or performance management
  • Credit scoring and insurance risk assessment
  • AI in critical infrastructure management
  • Systems used in education for assessment or admissions
  • AI in law enforcement or border control contexts

From August 2026, deployers of high-risk systems must maintain technical documentation, implement quality management systems, conduct conformity assessments, and register their systems in the EU database.

Limited Risk (Transparency)

Chatbots and AI-generated content fall here. The core obligation is disclosure: users must be told they are interacting with AI, and AI-generated content must be labelled. This is relatively straightforward to implement but often overlooked.

Minimal Risk (No Obligations)

Spam filters, AI in video games, and most internal analytics tools. No specific compliance requirements, though voluntary codes of conduct are encouraged.

The big compliance milestone arrives on 2 August 2026. From that date, providers and deployers of high-risk AI systems must have:

  • Risk management systems documented and operational
  • Data governance procedures covering training data quality, bias testing, and data retention
  • Technical documentation that describes the system's purpose, capabilities, and limitations
  • Record-keeping with automatic logging of system operations
  • Human oversight mechanisms allowing meaningful human intervention
  • Accuracy, robustness, and cybersecurity standards met and verified

General-purpose AI models (think GPT-4, Claude, Gemini) also face new transparency obligations from this date, including publishing training methodology summaries and complying with EU copyright law.

The UK has taken a different path. Rather than a single comprehensive AI law, the government has opted for a sector-specific, principles-based approach. The five core principles, published in 2023 and still guiding policy, are:

  1. Safety, security, and robustness
  2. Appropriate transparency and explainability
  3. Fairness
  4. Accountability and governance
  5. Contestability and redress

Existing regulators like the FCA, ICO, Ofcom, and CMA are expected to apply these principles within their existing powers. There is no single AI regulator and no equivalent to the EU's risk classification system.

For UK businesses, this creates a dual compliance challenge. You need to satisfy UK sector regulators and the EU AI Act if you operate across the Channel.

Here is what to do in the next four months before the August deadline:

1. Audit Your AI Systems

Map every AI system in your organisation. Include third-party tools, embedded models in SaaS platforms, and any automated decision-making. Most businesses discover they have three to five times more AI touchpoints than they expected.

2. Classify Risk Levels

For each system, determine whether it falls into the high-risk, limited-risk, or minimal-risk category under the EU framework. Pay particular attention to anything involving personal data, employment decisions, or financial assessments.

3. Assign Executive Accountability

The Act requires identifiable human responsibility. Appoint someone at board level or senior management who owns AI compliance. This is not optional and should not be buried in the IT department.

4. Document Everything

Technical documentation is the backbone of compliance. For high-risk systems, you need clear records of training data, model architecture, testing results, known limitations, and deployment parameters.

5. Implement Human Oversight

Every high-risk system needs a mechanism for meaningful human intervention. This means genuine override capability, not a rubber-stamp review that adds no value.

6. Review Your Supply Chain

If you use AI systems provided by third parties, you still carry compliance obligations as a deployer. Ensure your vendors can provide the documentation and transparency you need. Build this into procurement contracts now.

Fines under the EU AI Act scale with severity:

  • Prohibited AI practices: Up to 35 million euros or 7% of global annual turnover
  • High-risk system violations: Up to 15 million euros or 3% of turnover
  • Incorrect information to authorities: Up to 7.5 million euros or 1% of turnover

For UK SMEs, even the lowest tier is significant. But the reputational risk matters more. Clients increasingly ask about AI governance in procurement processes. Having a clear compliance posture is becoming a competitive advantage.

Compliance is not just a legal exercise. It forces good practice. Documenting your AI systems, testing for bias, implementing human oversight, and maintaining transparency are all things you should be doing anyway. The EU AI Act provides the structure and the deadline.

UK businesses that treat this as a box-ticking exercise will struggle. Those that integrate compliance into their AI governance framework will find it strengthens trust with customers, simplifies vendor management, and reduces operational risk.