AI can do the work. It cannot be accountable for the result

Dario Amodei, chief executive of Anthropic, has warned that artificial intelligence could bring “unusually painful” disruption to the labour market as a general substitute for human labour. For any organisation already using AI, a more immediate question matters more. When a machine helps produce the work, can anyone trust it, and who answers for it when it gets things wrong? That is a question about trust, and the people who manage it should sit where AI decisions are made. Too often, they do not.

 

 

A pledge that became a playbook

In 2025, more than 20 industry bodies, the Chartered Institute of Public Relations among them, signed the Venice Pledge, seven principles for the responsible use of AI. A year on, the CIPR has turned them into practical guidance, and their application extends well beyond the core functions of public relations.

 

The question is no longer whether to use AI, but how to use it ethically without sacrificing the trust businesses run on.

 

 

Why the stakes keep rising

Three forces are pushing AI ethics up the agenda. The first is reach. AI now sits inside everyday tools, often switched on by a supplier rather than chosen, so few leaders can quickly identify where it already touches their customers.

 

The second is legislation. As the EU AI Act’s transparency rules phase in, businesses will need to label AI-generated content and disclose deepfakes, with the heaviest obligations falling on synthetic media. Any company reaching European audiences is already in scope, and similar laws are expected in the UK soon.

 

The third, and most concrete, is liability. In 2024, a Canadian tribunal held Air Canada liable after a chatbot gave a customer wrong information. The airline argued the bot was responsible for itself and lost. Every board should absorb the lesson. A company owns what its AI tells the public.

 

 

How to embed responsible AI

Responsible AI is less about lofty principles and more about a few prosaic routines for any sector. Start by knowing what you are running. Keep a register of where AI is used, including bought-in tools, and who owns each one. Make a named person accountable for anything that reaches a customer, employee or investor, because accountability cannot be handed to software. Match oversight to the stakes. A marketing caption needs a lighter touch than a system screening job applicants or approving payments. Build verification into the schedule, since AI can be confidently wrong, with a second check for anything involving money or legal rights. Invest in people.

 

The skill that protects a business is not writing a clever prompt; it is knowing when the answer is wrong. Most importantly of all, make sure that the people who are exercising the oversight know enough to understand when AI has got things wrong – ‘human oversight’ requires active critical engagement with AI drafts and is not just a formal sign-off protocol. 

 

 

What businesses can learn from communications

The standards of whoever produces a company’s external messaging become the standards the public judges the whole firm by, and they apply equally to an AI-written email, a chatbot quoting a returns policy, or the figures in a board paper. It is why the Pledge calls for an Active Global Voice. Communicators are too often absent from AI governance, where their judgement is needed most.

 

Here, awareness, openness and transparency do the heavy lifting, and the CIPR’s guidance turns it into a practical test for when to disclose. If AI produced the substance and a person reviewed it, say so where the audience would assume a human wrote it. If a person did the work and AI merely tidied it, the human remains the author and no disclosure is needed. If AI runs quietly in the background, a line in your published AI policy is enough.

 

In 2026 research by Meltwater and YouGov, 86 per cent of consumers said brands should disclose AI-generated content, and around a third trust a brand less when they find it was used and not declared. What matters is not whether AI was involved, but whether a person stands behind the result.

 

Amodei is right that AI will reshape how the work gets done, but it cannot carry responsibility for the outcome. That belongs to people, which is why the team that best understands trust should be in the room where AI decisions are made. In a market awash with cheap, automated output, the valuable thing is work someone will put their name to, and that trust is what earns a business its licence to operate. 

 

---

 

Alastair McCapra is CEO of the Chartered Institute of Public Relations

 

Main image courtesy of iStockPhoto.com and Natee127