AI in Criminal Law: Practical Uses Lawyers Can Deploy Today

Kristina Kothrakis, Director at Doogue + George Defence Lawyers, takes a middle path. For her, AI is already an assistant with very tangible uses — one that lawyers can and should be experimenting with now.

“AI isn’t about replacing lawyers,” she says. “It’s about giving us a platform to work from, saving hours of background work, and letting us focus on strategy and advocacy.”

Earlier this year, Ms Kothrakis spoke at the International Bar Association European Fraud and Compliance Lawyers Conference in Paris on the intersection of AI and criminal practice. She has since continued that conversation back home, sharing practical insights on how these tools are transforming the day-to-day workflow of defence lawyers.

Chronologies in White-Collar and Complex Matters

The clearest example comes from the world of Commonwealth white-collar prosecutions.

“These cases often involve literally thousands of dates, transactions, communications, and corporate records. Building a chronology out of that volume of material can take weeks of manual collation. It’s a vital step, but an incredibly resource-intensive one,” Ms Kothrakis explains.

Large Language Models (LLMs) change that equation. By processing the raw documents, they can generate a structured chronology in minutes. “Of course, no lawyer should accept that output blindly. But instead of starting with a mountain of disconnected data, you begin with a scaffold that already pulls the evidence into order. Then it becomes a process of checking, refining, and strategically shaping it.”

The cost implications are significant. What previously consumed endless junior hours can now be condensed into a manageable review process. That efficiency not only benefits clients financially but also allows defence teams to reach substantive strategy discussions earlier in a matter.

Spotting Inconsistencies Across Witnesses

Another powerful application is in detecting inconsistencies — particularly where multiple witnesses describe the same critical event.

“We had a case where a large number of witnesses were all talking about the same incident. Their statements overlapped in many places, but it was the differences that really mattered for our cross-examination strategy. Running those statements through an LLM produced a ready reckoner: a single comparative map of what each person had said.”

This doesn’t replace legal judgment, but it dramatically accelerates it. “Instead of trawling through dozens of statements manually, we could see at a glance where accounts diverged. That gave us more time to think about which inconsistencies mattered, and how to put them to witnesses in a way that the court would find persuasive.”

For busy criminal practices, this kind of tool isn’t just convenient — it’s transformative.

Marshalling Notes and Specialist Resources

AI isn’t only useful for processing formal evidence. It can also be deployed internally, to help lawyers manage their own knowledge bases.

“I keep extensive specialisation notes on criminal law. Keeping them up to date, and reorganising them into a usable form, is essential but tedious,” Ms Kothrakis notes. “AI makes it easier to marshal those notes into a living, searchable resource. You can ask questions of your own material and get it back in a clean, court-ready form.”

This kind of internal application avoids many of the confidentiality pitfalls of external data processing and is particularly powerful for practitioners who need a constantly updated toolkit.

Security and Confidentiality — The First Priority

But for Ms Kothrakis, no discussion of AI in practice can proceed without addressing the elephant in the room: security.

“Confidentiality is the lifeblood of the solicitor-client relationship. You cannot responsibly use AI without being absolutely satisfied about how information is handled.”

That means understanding whether data uploaded to a platform will be stored, whether it will be used to train future models, and whether the system complies with local data-protection requirements.

“In practice, that often means avoiding public AI platforms and instead working with secure, jurisdiction-appropriate tools. Lawyers need to know where their client data is going, and regulators will expect nothing less.”

She warns against a casual approach: “Unchecked, AI risks turning legal shortcuts into legal pitfalls”

Balancing Efficiency with Judgment

The broader message is one of balance. AI offers speed, efficiency, and cost-savings — but not judgment.

“The lawyer’s role doesn’t disappear. These tools are platforms, not substitutes. They can highlight inconsistencies, assemble chronologies, or restructure notes. But it’s the lawyer who decides what matters, how it fits into the law, and how it should be put in front of a court.”

That combination — efficiency plus judgment — is where the real value lies. Clients benefit from reduced costs and faster turnaround, while lawyers gain back precious time for high-level analysis, negotiation, and advocacy.

Lessons from Paris: A Global Perspective

At the IBA (EFCL)conference in Paris, Ms Kothrakis saw that these issues are not unique to Australia.

“Across jurisdictions, the same themes emerged: everyone is excited about the potential of AI to save time, but everyone is also deeply conscious of security and confidentiality. Regulators in Europe and North America are already scrutinising this closely. The consensus was that lawyers can’t afford to ignore AI — but they also can’t afford to adopt it recklessly.”

Her take-away is that the Australian profession needs to start building comfort now, rather than waiting. “If we treat AI as an experiment in carefully bounded tasks, we’ll learn how to use it wisely. If we bury our heads in the sand, we risk being overtaken by events.”

Looking Forward

Ms Kothrakis is clear-eyed about the limitations of AI. “It’s not a magic wand. It won’t write your submissions or win your cases. But it will save you hours of repetitive work. And if you’re a criminal lawyer with a diary full of trials, that’s no small thing.”

She encourages colleagues to start small: “Try it on a chronology. Try it on marshalling your own notes. See how it can surface inconsistencies across documents. Always check and refine — but don’t be afraid to test its limits.”

Her closing reflection is simple but powerful:

“AI is not the future of criminal law. It’s already part of the present. The firms that use it wisely — with care for confidentiality, and always with a lawyer’s judgment — will free themselves to do what they do best: think, argue, and advocate.”

Pull-Quotes for Layout:

• “Unchecked, AI risks turning legal shortcuts into legal pitfalls”

• “Instead of starting with a mountain of disconnected data, you begin with a scaffold that already pulls the evidence into order.”

• “The firms that use it wisely will free themselves to do what they do best: think, argue, and advocate.”