Last weekend, at the 5th International Ankara Brand Meetings organized by the Ankara Chamber of Commerce, Vircon Legal was represented by our Managing Partner Erdem Mümtaz Hacıpaşaoğlu. On a panel devoted to the relationship between law and artificial intelligence, Hacıpaşaoğlu took the stage alongside Cem Şanap, Orhan Fahri Başçavuşoğlu and Serdar Ülkü. The Q&A ran past the session’s allotted time, drawing exceptionally strong engagement.
The central thesis of both the panel and the most frequent off-stage question — “so what does the legal profession look like in this new world?” — was clear: AI simultaneously collapses three foundational assumptions of law: that violations are costly, that deterrence works, and that liability can be attributed to a single actor; in a world where friction has been zeroed out, both legal risk layers and the form of the profession need to be redesigned.
I. The legal ground beneath AI: “do first, answer if asked”
Hacıpaşaoğlu opened by reminding the room that today’s large language models are built on a long series of legal stretches — copyrighted books, unlicensed images, scrapes that ignore robots.txt. The industry’s story is largely the story of “do it first, answer if asked” — or, in the more familiar formulation, “it’s easier to ask forgiveness than permission.” Getty v. Stability AI, NYT v. OpenAI, the Authors Guild case, Sarah Silverman’s lawsuit, GitHub Copilot — all are circling the same foundational question: is unauthorized learning infringement? Even the 2011 Naruto v. Slater (monkey selfie) case touches in surprising ways on today’s “who owns AI output?” debate.
One prompt away: friction has been zeroed out
The turning point of the talk was this observation: we are now one prompt away from a legal violation. In the past, imitation, forgery or identity exploitation required expertise, equipment, time, and often an accomplice; this friction was law’s indirect protective layer. Today, a single prompt can produce deepfakes, voice clones, unlicensed images, and text in someone else’s style. Hacıpaşaoğlu underlined that this collapses three foundational assumptions of law at once: that violations are costly, that deterrence works, and that liability can be attributed to a single actor. Is the prompter liable? The model builder? The platform? The data provider? The classical toolkit cannot keep pace with this speed and dispersion.
The conclusion was a single clear sentence: we will not refrain from using these tools — we should use them. But we must do so knowing that the training data, the generated output, and the user input are each their own separate legal risk layer.
II. So how does the legal profession continue in this world?
This was the most frequent off-stage question. Hacıpaşaoğlu’s answer runs through transformations Vircon Legal has been implementing in its own practice over the recent period.
The client relationship has changed
In the past, lawyers often answered questions arriving from other lawyers or from colleagues in different specializations. Today those questions arrive from Claude, Gemini and OpenAI; clients often explicitly state which model is doing the asking. Hacıpaşaoğlu noted that there is no point fighting this — and that “AI says X, what do you say?” has become an increasingly high-quality intellectual exercise.
The pricing model has been rebuilt
Law firms largely run on hourly billing. According to Hacıpaşaoğlu, rather than racing against AI, the right move is to rebuild pricing and team structure around it. There were two roads: either say “AI is increasing our productivity” and raise hourly rates, or lower prices to keep legal services accessible. Vircon Legal chose the second: the retainer model — which Vircon Legal has called “lawyer as a service” for years — was repriced to remain accessible especially for early-stage entrepreneurs. Every workstream that does not need to be project-based was moved into this model.
AI usage: all the way in, but hybrid
Operating its own language model since the Vircon Group era, the team has watched capabilities grow alongside the risks. Hacıpaşaoğlu noted that, due to client confidentiality, trade secrets and personal data, local LLMs are now used inside a hybrid architecture: any client data is first processed locally; where parameter capacity falls short of required competence, pseudonymization is applied locally before anything leaves the perimeter. In short: no client data leaves with information that can identify the client.
Teams are getting smaller, focus is deepening
Hacıpaşaoğlu observed that legal teams are shrinking and that Vircon Legal is following the same path. He sees this not as a loss but as a sharpening: fewer people, deeper specialization, less work but more accurate work. He expects the “solopreneur” movement to appear in legal practice as well.
Closing: the legal profession in its current form will not survive
Hacıpaşaoğlu’s candid forecast: the market is shrinking, the payments balance is deteriorating, and the legal profession in its current form will not survive. But what replaces it is not necessarily bad — more specialized, less crowded, and coordinated with AI. He noted that Vircon Legal has chosen to make this transition early, and thanked the Ankara Chamber of Commerce and Cem Şanap for the invitation.
Highlights from this panel
- Legal ground: “Do first, answer if asked” and the resulting lawsuit chain
- Zeroed-out friction: Violation one prompt away, three collapsed assumptions
- Risk layers: Training data + generated output + user input
- Client relationship: “AI says X, what do you say?” as a new question
- Pricing model: Repricing lawyer-as-a-service
- Hybrid AI: Local-first processing and pseudonymization
- Smaller teams: The solopreneur trend coming to legal practice
You can read the full newsletter on LinkedIn.
