On 16 June, Bernima's CEO Dr. Sergey Milutin joined the AI2L London Summit at Mishcon de Reya's offices on Kingsway — an afternoon-to-evening gathering of general counsel, legal innovators, and technology builders convened to examine where legal AI is actually heading.
AI2L — the International Alliance of Legal Innovation, Lawtech and Legal AI — is a global series convened by the legal-tech associations of Brazil (AB2L), Europe (ELTA), Portugal (APLT), India (ILTN), and Asia-Pacific (ALITA), in partnership with LegalTechTalk. That structure gives the London Summit a cross-jurisdictional character: not a single market talking to itself, but practitioners from very different legal systems comparing what works in production.
For a company building secure infrastructure for legal AI, it was a high-signal room.
The agenda has moved from capability to control
What stood out across the programme was how little time was spent debating whether AI belongs in legal work. That question is settled. The harder conversations were about governance, risk, and trust — the conditions under which a regulated organisation will actually run AI against privileged work.
The throughline was consistent. The institutions furthest along are not the ones with the most capable models. They are the ones who can answer the control questions: who is accountable, where the data resides, how outputs are verified, and whether the workflow is defensible after the fact.
Why this maps directly onto Bernima's product thesis
This is the conviction Bernima is built on. Legal AI that cannot evidence security, confidentiality, and a defensible record does not get adopted by the firms and legal departments that matter — it gets piloted, then quietly shelved. The pilot-to-deployment gap in legal AI is not a model-quality problem. It is a control problem.
That gap is the product. In legal work, raw model capability is necessary but not sufficient; what converts capability into deployment is the system around the model — control over where data lives, assurance that privileged matter information stays contained, and a record a general counsel can stand behind in front of a regulator, a board, or a client. Bernima is built to own that layer rather than treat it as an afterthought.
It was useful to hear the buyer's requirements stated so plainly. When the decision-maker is a GC managing risk across dozens of entities, the criteria are concrete: data residency, access control, auditability, and the ability to prove what the system did and why. These are the requirements that decide enterprise legal deals — and the ones Bernima designs around from the first line, rather than retrofitting once a deal stalls on security review.
A genuinely cross-border demand signal
A session on AI-powered dispute resolution drew on perspectives from Europe, India, Asia-Pacific, Africa, and Brazil, making plain how differently legal innovation is unfolding by market even as the requirements for trust and governance converge.
For Bernima that convergence matters commercially. Regulatory specifics differ by jurisdiction, but the underlying demand is consistent: secure, governable AI that lawyers can rely on under their own confidentiality obligations. Hearing it articulated by in-house leaders across several continents in a single afternoon is a useful demand signal for a category that is still forming.
What we took away
The Bernima team left the AI2L London Summit with new relationships across the international legal-innovation community and confirmation of the thesis we are executing against. The market has matured past novelty. The people who hold the budget are asking about governance, accountability, and control — and they are asking hard.
Those are the questions Bernima is built to answer.
Bernima is building secure infrastructure for legal AI — software and a dedicated secure environment for law firms and legal departments. Contact us to schedule a call with the Bernima team, learn more about the product, and request a demo.