When AI Reads Everything: Rethinking Privilege in the Digital Age

---

The volume of digital information entering disputes is growing rapidly with voice notes, AI-generated meeting notes, chatbot conversations, and even prompts entered into internal AI tools. As the boundaries of disclosure continue to expand, so too do the challenges around confidentiality and legal privilege.

Punam Mehta, Group General Counsel at Longevity Partners, recently sat down with Future Disputes UK for an exclusive Q&A to share her insights on how AI is reshaping disclosure, document review, and dispute management. 

She also addresses responsibility when sensitive dispute information is exposed through technology and stresses the importance of protecting privilege proactively.

Enjoy the interview below.

---

1. Has the growing amount of digital information changed the nature of disputes themselves? Are disputes becoming more complicated because there are now so many emails, documents, data records, and other digital evidence?

Yes - and it happened quickly. The typical documents at the disclosure stage of a dispute have evolved from letters, to emails, to now WhatsApp and Teams messages, voice notes, and other forms of digital evidence. Each format brings its own volume and complexity - a chat thread doesn't read like a formal letter, and working out what was actually agreed or intended is a skill, and can often be interpreted differently. AI inevitably adds another layer to that. Chatbot conversations, AI-generated meeting notes, and even the prompts fed into internal tools may all become disclosable in the same way emails are now. The direction of travel is more formats, and volume - which makes the disclosure exercise itself a much bigger part of running a dispute than it used to be.

2. Do you think AI will mainly become a powerful assistant to lawyers and dispute professionals, or will it eventually take responsibility for some decisions currently made by humans?

Disputes are document and evidence heavy, so AI's biggest impact so far has been as an assistant - trawling through huge data sets, identifying trends, running keyword searches, and producing chronologies that until recently would take trainee and junior lawyers hours, sometimes days, to complete. Of course there is an efficiency gain here, and it frees up time for the more judgement-based parts of the job.

The inevitable risk of over-reliance is that you may miss something which is a smoking gun - AI is only as good as the prompt and the data set it's working from, and it can miss context or nuance that a trained eye would catch. I don't think it's likely to take responsibility for the strategic decisions made in a dispute - whether to settle, how to frame a case, what risk to take on - those still require human judgment, commercial awareness, and accountability. However, it may increasingly be used to decide the next step of a document review exercise, depending on the scope of the initial prompt provided to it. Over time, that could shift where the line sits between assistant and decision-maker, even if the strategic calls stay firmly human.

3. In your opinion, who should bear responsibility if confidential dispute information is exposed because of a technology platform or AI system?

It depends on the context. Presuming it's a data breach, the primary responsibility is likely to reside with the platform or system provider - they've taken on the obligation to keep that data secure, and if their systems fail, that's on them. That said, it depends on whether the information in question should have been on the platform in the first place, i.e. the details of a confidentiality agreement. If a party has uploaded highly sensitive dispute material to a tool that wasn't approved or covered by the relevant confidentiality terms, responsibility shifts back to them for that choice. In practice, I think this will be tested through the contracts themselves - who warranted what, what due diligence was done on the platform, and what data-handling obligations were actually in place before the breach happened. Responsibility is rarely going to sit in one place alone.

4. Could online dispute resolution become the normal first step for certain types of commercial disputes? From your perspective, what types of disputes are most suitable for this model, and which are not?

I think it could become the norm, particularly for disputes that are lower value, more binary in nature, or where the facts are largely undisputed and it's a question of applying a formula or a clear contractual mechanism - things like straightforward payment disputes, consumer complaints, or disagreements over set fees or standard terms. Online dispute resolution works well when there isn't a huge amount of nuance to argue over and the parties want a fast, proportionate outcome rather than a drawn-out process.

I don't see disputes involving complex facts, multiple parties, cross-border elements, or reputational sensitivity or anything involving detailed technical or expert evidence being part of such a model. Those need a level of nuance, cross-examination, and human judgement that an online-first model isn't yet built to handle well. My sense is we'll end up with a tiered system - online resolution as a genuine first step for high-volume, lower-complexity disputes, with traditional routes reserved for anything that needs deeper scrutiny.

5. Do you think the future will bring fewer disputes because technology can identify problems earlier, or more disputes because technology makes it cheaper and easier to challenge decisions?

We're already seeing an increase in consumer disputes as a result of AI, and that's a likely continuing trend - people are more aware of their rights, more willing to challenge a decision, and AI tools make it easier for them to do so cheaply. In the B2B commercial sphere, it's likely to be more of a mix. AI will facilitate access to justice for many and may bring more clarity and speed - smaller businesses that couldn't previously afford extensive legal support may now be able to build a stronger case more cheaply. At the same time, it's creating its own complexity, because new problems are being created by this technology and by how we use it - e.g. disputes over AI-generated content, AI decision-making, and liability for AI errors are only going to grow - and new types of privilege issues. 

We may also end up in an arena where the calibre or size of a law firm matters less than it used to, if AI is used to prepare documents that historically required a lot of time and skill to get right. That could level the playing field in some respects, but it will also put more pressure on judgment and strategy as the real differentiators, rather than sheer resource. Taking that further, we may end up with something like litigation between advanced AI models - systems predicting next steps and strategy that are then checked and administered by humans. I hope not, because that would be a fairly dystopian outcome, but it's not impossible.

Whatever the future looks like, I don't think AI in itself will bring fewer disputes, because disputes arise for mainly human reasons. AI used to meet operational challenges, ensure compliance, steer behavioural change, and assist with sound decision-making may result in fewer disputes - but even that outcome depends entirely on how humans choose to use it.

6. What do you expect the audience will take away from your session at Future Disputes UK?

I'd like the audience to leave with a better understanding - or, honestly, more useful questions - about the increasingly complex environment we now operate in, particularly if you're part of a global company using AI in your day-to-day work. Different jurisdictions treat privilege, disclosure, and AI-generated material differently, and that complexity only grows the more tools and platforms you introduce into your workflow.

Beyond that, I want people to walk away with practical steps on how to build the infrastructure to safeguard privilege as a precaution, rather than scrambling to fix it once an issue has already arisen. That means thinking now about how AI tools are procured, what data they can access, how conversations and outputs are labelled and stored, and who within the business is responsible for that oversight. The organisations that get ahead of this now will be in a much stronger position than those that wait until a dispute forces the issue.

---

As Punam highlighted, protecting privilege in an AI-enabled environment starts well before a dispute arises. Clear controls around which tools can access sensitive information, how data is stored and labelled, and who is responsible for oversight can help organisations stay ahead of increasingly complex disclosure and confidentiality risks.

This September, join Punam at the fireside chat "The Privilege Crisis: When AI Reads Everything", where she will explore what happens to legal privilege when AI tools process confidential communications, e-discovery platforms scan vast amounts of information, and AI becomes part of the legal workflow.

Register now to join the discussion at Future Disputes UK.