"At PwC, we take risk seriously in every aspect of our work because the trust placed in us depends on it. In this year’s Annual Review, we wanted to shine a light on how that commitment is evolving in the age of Generative AI. The opportunity presented by AI is significant, but so is our responsibility to use it well. By grounding innovation in clear principles, strong governance, and professional judgement, we can adopt AI with confidence while protecting the quality and integrity that our clients and our people expect from PwC."
Ethics in the age of AI
Generative AI has moved from experimentation to everyday use across our firm. It drafts, analyses, summarises, and accelerates, and with that opportunity comes a clear obligation: to use it in a way that is lawful, ethical, transparent, and accountable. AI governance is not a brake on innovation. It is the framework that makes confident, large-scale adoption possible, protecting the trust our clients, our people, and society place in us.
This year we saw that framework take firm shape. The PwC Network AI Policy was approved by the Global Board Risk Committee, setting out consistent principles for how AI is developed, deployed and used across PwC firms and global functions. It gives us a common foundation for managing AI responsibly and mitigating risk, supported by the Global AI Business Rules and, where applicable, more rigorous local requirements.
"AI can enhance how we analyse information, develop insights, and deliver our work, but it does not replace human accountability. Every output still requires informed challenge, careful validation and sound professional judgement. Our responsibility is to ensure that greater speed and capability never come at the expense of accuracy, transparency or quality."
Seven principles that guide how we use AI
The Network AI Policy defines seven core principles. Together they translate our values into practical expectations for anyone using AI in their work.
Quality does not change with the tools we use
The quality and integrity of everything we do is our responsibility, personal to every partner and member of staff, and unchanged by how the work is produced.
Our expectations are therefore unambiguous. Human judgement and review are not optional and cannot be delegated to a tool. Facts, sources, citations and conclusions must be validated with the same discipline we bring to every aspect of our work, and the level of scrutiny should reflect the visibility and risk of the content. Where standards are not met, our existing performance and accountability frameworks apply, including where quality was compromised through reliance on AI.
From principles to daily practice
Principles only matter if they are lived. Through the year we translated the policy into practical guardrails for day-to-day work:
| Expectation | What we ask of our people | |
| Approved tools only | Generative AI is used only through firm-approved and governed platforms, so that data stays within our protected environment. | |
| No unprotected confidential data | Client, personal, or otherwise confidential information is not entered into unapproved or public AI tools. | |
| Human review before issue | Every AI-assisted output is reviewed by a suitably qualified person before it is issued, delivered, or published. | |
| Verify facts and sources | Facts, figures, sources, citations, and conclusions are validated with the same discipline applied to all our work. | |
| Proportionate scrutiny | The depth of review reflects the nature, visibility, and risk of the content, with client-facing and public material warranting particular care. | |
| Client contracting | AI-related contractual terms are agreed in line with the Network AI Contracting Guidance, with Office of General Counsel (OGC) and Risk and Quality (R&Q) support where needed. |
"Responsible AI is achieved through the choices we make every day: using approved tools, protecting confidential information, verifying what AI produces, and speaking up when something does not look right. When our people understand both the possibilities and the boundaries, we create a culture in which innovation can scale safely and responsibly."
Building capability and a culture of responsible use
Responsible use depends on informed people. Our participation in the Network AI Academy connected our teams to a community that now spans thousands of AI Champions across the network, with regular champion calls, Copilot Power Hours, and peer-led AI Clubhouses covering everything from methodology and guardrails to citizen development and governed low-code solutions. Locally, these sessions have helped colleagues understand not only what the tools can do, but where the boundaries sit and why they exist.
We have paired this with clear communication of our expectations: use approved tools, protect confidential information, disclose AI use where appropriate, and never let speed substitute for judgement. Training integrity applies here too, learning and assessments are to be completed honestly and independently, whatever assistance is available.
Oversight and what comes next
AI governance is an ongoing discipline rather than a one-off exercise. Monitoring, incident reporting and consultation routes give us visibility of how AI is being used and where risks are emerging and allow us to respond collectively as a network. Further network guidance, tools to help identify errors in AI-assisted work, and enhanced monitoring capabilities are in development, and we will continue to align our local policies and review processes as they arrive.
By anchoring our use of Generative AI in clear principles, approved technology, rigorous human review, and open accountability, we are able to embrace the opportunity AI offers while holding firm to the standard that defines us: doing the right thing, in everything that carries the PwC name.
Contact us
