Generative AI has moved rapidly from novelty to necessity. In boardrooms, legal teams, product groups and C-suites alike, AI tools are now routinely used to summarise documents, stress-test arguments, draft communications and organise complex thinking. For many professionals, AI has become a cognitive copilot – faster than search, broader than memory, and always available.
But as organisations increasingly turn to AI for high-stakes decision-making, a critical question is coming into sharper focus: where do longstanding legal principles fit in a world of machine-generated insight?
One area where this tension is particularly acute is legal professional privilege – a cornerstone of common-law legal systems (such as the US and English legal systems) that enables clients to seek legal advice (or prepare for litigation) in confidence.
Recent court decisions on both sides of the Atlantic underline a simple but often overlooked truth: AI may transform how we process information, but it remains subject to longstanding legal principles of privilege.
Privilege was built for trust – not tools
In England and Wales, legal professional privilege (LPP) protects confidential communications made for the purpose of obtaining legal advice or conducting litigation. It has been described by the courts as a “fundamental condition” underpinning the administration of justice, and a “fundamental human right”.
But privilege has boundaries. It depends on three core elements:
- a defined legal purpose (advice or litigation)
- in the case of legal advice privilege, a lawyer being involved
- and confidentiality
AI challenges each of these in different ways.
No matter how sophisticated a chatbot may appear, it is not a lawyer. Communication between a non-lawyer and an AI system therefore fall outside legal advice privilege entirely. Even enterprise-grade, closed-system AI tools – while addressing confidentiality concerns – cannot on their own satisfy the “lawyer in the loop” requirement that legal advice privilege demands.
As a judge put it in a recent US case, privilege requires “a trusting human relationship”. In short, automation does not replace professional status, and intelligence – artificial or otherwise – is not the same as legal authority.
Confidentiality: the hidden fault line
Confidentiality is where many organisations inadvertently create risk.
Publicly available AI platforms are typically provided on terms that allow user inputs to be stored, analysed or used for model training. Even where there is no intention to disclose information, the absence of robust confidentiality protections can be fatal to a claim of privilege.
Courts have increasingly recognised this reality. In a recent UK Upper Tribunal decision, the judge observed that placing client material into an open-source AI tool amounts to putting it into the public domain – breaching confidentiality and waiving privilege.
A US federal court reached a similar conclusion when considering whether documents generated through a commercial AI platform were protected by attorney-client privilege. They were not – in part because the terms of use allowed disclosure to other parties.
Further decisions are expected to come thick and fast, and the analysis is likely to become more nuanced, with an impact on other established aspects of litigation. For example, in Morgan v. V2X, Inc., the US District Court for the District of Colorado held that AI-assisted litigation materials prepared using public AI tools were protected from disclosure.
However, it also prohibited material designated as confidential under a protective order from being uploaded into consumer AI platforms such as ChatGPT or Claude, unless the provider was contractually barred from training on inputs or disclosing them to third parties. Enterprise AI with negotiated safeguards was permitted; consumer AI was not.
When AI is used before the lawyer arrives
One of the most consequential – and least discussed – risks arises before lawyers are involved at all.
Many employees and executives now use AI tools to structure their thoughts, draft internal papers, prepare initial instructions for legal teams, or ‘stress-test’ proposed legal strategies. While often well intentioned, this “AI-first” approach can mean that sensitive legal issues are explored in environments where privilege never arises.
Once created, those materials do not become privileged simply because they are later shared with a lawyer. As one US judge put it, documents generated through AI are not “alchemically transformed” into privileged material after the fact.
Enterprise AI doesn’t eliminate governance risk
Enterprise GenAI platforms go a long way towards addressing data security and confidentiality concerns. But they are not a complete solution.
Where non-lawyers use AI to obtain what is effectively legal advice, privilege may still never arise – regardless of how secure the system is. This is not a technical failure; it is a governance one.
The lesson is clear: AI strategy and legal strategy cannot be developed in isolation.
A moment for organisational recalibration
The rapid adoption of GenAI has created new pressure points, but it also offers an opportunity.
Forward-thinking organisations are using this moment to revisit internal policies on confidentiality, data handling and AI usage – not to restrict innovation, but to channel it responsibly. Clear guidance on when legal teams should be engaged, and on how AI tools can and cannot be used for legal matters, is becoming a core element of AI governance.
Crucially, this is not about resisting technology. Lawyers themselves are increasingly using AI as part of their working processes, where protections such as litigation privilege and the “working papers” doctrine may still apply. The distinction lies in who is using the tool, for what purpose, and within what safeguards.
Old principles, new technology
Generative AI may well revolutionise how organisations analyse information and make decisions. But its novelty does not place it outside the law.
As one judge recently observed, while time will tell whether AI fulfils its transformative promise, longstanding legal principles still apply. Trust, confidentiality and human responsibility remain central – even in an age of intelligent machines.
For leaders navigating AI adoption, the message is not to slow down, but to be deliberate. The most successful AI strategies will be those that combine technological capability with clear human oversight – ensuring innovation does not come at the expense of legal certainty.
