Beyond the watermark: What the EU AI Act means for CX.

When Anthropic announced that Claude would start watermarking its AI-generated content to comply with the EU AI Act, the questions followed almost immediately.

• Do we need to start watermarking our own content?
• How does this work?
• What are our responsibilities?

It's a hot topic in the AI world right now, but answering the watermarking question is the easy part. Understanding what the Act truly means for customer-facing AI is where CX leaders need to focus.

Provider vs. deployer: who is responsible for what?

Under the EU AI Act, providers of AI systems that generate synthetic content — including text, audio, images and video — are required to ensure their outputs are marked in a machine-readable format and detectable as artificially generated. This is why Anthropic tags Claude’s outputs.

However, most customer-facing organizations act as deployers, which carries a separate set of responsibilities:

• The human review exemption: For AI-generated or manipulated text, the deployer disclosure duty applies where the text is published with the purpose of informing the public on matters of public interest. Within that scope, disclosure is not required where the content has undergone human review or editorial control and a person holds editorial responsibility. For example, if your communications team uses an AI tool to draft a public statement on a data-breach incident and an editor substantially revises the content, verifies accuracy and takes editorial responsibility for the final publication, the disclosure obligation would not apply.
• What doesn't exempt you: Basic grammar or spell-checking does not count as meaningful human review in the context of the watermarking rules. If the core text remains unchanged, the obligation to disclose AI generation remains.
• The global reach: The Act functions with broad extraterritorial reach. It applies to providers placing an AI system on the EU market, deployers established or located in the EU, and providers or deployers established or located outside the EU where the output of the AI system is used in the EU — regardless of where your company is headquartered.

The CX obligations that go beyond watermarking

Watermarking is the headline, but it's not the only transparency duty that touches customer experience. A few others worth knowing:

• Chatbot transparency. When a customer is interacting with an AI system rather than a person, they need to be told. This applies to limited-risk systems like chatbots and virtual assistants.
• Emotion inference. The Act restricts inferring emotions in ways that could be manipulative. CX tools that analyse tone, sentiment or biometric cues to steer a conversation need a careful look.
• Human oversight. Deployers must assign competent human oversight, monitor system operations and suspend use if there's a risk. For contact centres, that's often already in place — but it needs to be documented, not assumed.
• Recordkeeping. Operations must maintain robust system logs and document incident-reporting workflows.

3 compliance steps CX leaders should take now

The regulation is complex, but getting started doesn't have to be. Three practical steps will help put you on solid ground:

1. Assess where AI touches the customer. Map every customer-facing AI tool — chatbots, copilots, sentiment analysis, summarisation — and determine if it triggers a transparency or disclosure obligation.
2. Confirm your human review process is genuine. If you rely on the human review exemption, make sure the review is meaningful, not a rubber stamp. Document who holds editorial responsibility.
3. Build your record-keeping before someone asks for it. The Act is specific about logs, oversight and incident reporting. Getting this in place now is far easier than reconstructing it under pressure.

AI governance as an innovation enabler

Content watermarking is only one of the first milestones in broader wave of global AI regulation. But for CX leaders, the core lesson is clear: legal accountability falls on how your business deploys these tools, not just who built them.

Taking proactive steps toward compliance and transparency today will do more than mitigate regulatory risk. Clear governance creates an operational framework for CX teams to innovate rapidly and keep up with customer needs — all while protecting customer trust.

Governance isn't a bottleneck; it's your license to scale AI safely.