Every Firm I Talk to Has the Same Blind Spot
They have selected strong AI tools research platforms, drafting assistants, and general-purpose models, often more than one, and usually with considerable care. What almost none of them have done is ask a much harder question:
Who is accountable when one of these tools is wrong?
Not “wrong” in the abstract. Wrong in a filed brief. Wrong in a client email. Wrong in a way that a judge, a Bar Association, or a malpractice carrier eventually notices.
The real risk is not AI adoption itself. It is the gap between adopting AI and governing it.
The question is not whether AI works. The question is who is responsible when it does not. Every new tool changes the way work is done. It does not change the need for accountability. The failure of management begins where responsibility has no owner.
AI Adoption Is Fragmented
Most firms are no longer using a single AI solution. They are operating multiple platforms, each adopted for a different purpose and often by different teams.
An organization does not become more effective by acquiring more tools. It becomes more effective by making their work coherent.
Each system may perform its intended task well. Yet few firms have mapped how these tools interact, where they overlap, what data flows between them, or who is responsible for oversight when something goes wrong.
The capability exists. The governance often does not.
My Perspective: The Standard Remains the Standard
The first decision is which tool to buy.
The more important decision is who is accountable for its results.
An organization has not truly managed AI until it has clearly defined who is responsible for its outputs. The maxim remains unchanged:
Ignorance of the law is no excuse.
AI is no exception.
While many firms have treated AI as a procurement decision, regulators and courts have not.
The ABA Model Rules of Professional Conduct make clear that this is not new territory requiring new rules. It is existing professional responsibility applied to new technology.
The governing standards remain unchanged:
- Competence (Model Rule 1.1)
- Confidentiality and Data Security (Rule 1.6)
- Candor to the Court (Rule 3.3)
- Supervision of Non-Lawyer Assistance (Rule 5.3)
None of these obligations contain an AI exception.
Courts have already made the consequences of ignoring these responsibilities tangible.
In Mata v. Avianca, Inc. (S.D.N.Y. 2023), Judge Kevin Castel sanctioned attorney Steven Schwartz under Rule 11 after a filing contained multiple fabricated case citations generated by ChatGPT. The court also ordered corrective action regarding the false citations submitted.
That decision was issued in 2023.
The individual cases are no longer the story.
Courts continue to encounter AI-generated inaccuracies, fabricated citations, and unsupported legal authorities. Increasingly, the judiciary is treating these incidents not as isolated technology failures but as failures of professional supervision.
The court reiterated that while generative AI may be a new technology, the same professional and procedural standards still apply.
The response has extended beyond the courts.
Insurers no longer ask simply whether a firm uses AI. They ask which systems handle client information, how outputs are reviewed, and who is accountable for that review.
The question is no longer whether AI is being used.
The question is whether it is being governed.
An organization cannot govern what it does not understand. Responsibility begins with knowing where AI is used, how it is used, and who is accountable for its outcomes.
Turning AI Governance into Competitive Advantage
The solution is neither fewer AI tools nor a single tool for every task.
Both mistake the instrument for the problem.
The organizations that use AI effectively begin with a different question. They ask not, “Which tool should we use?” but rather, “Who is responsible for the work before it leaves the organization?”
That is a management decision, not a procurement decision.
Effective governance recognizes that not all work requires the same level of review.
Routine tasks may be automated.
Work that influences a client, a court, a regulator, or a significant business decision requires deliberate human judgment.
The measure of an AI strategy is not the sophistication of its tools.
It is the clarity of its accountability.
Every piece of work should have an identifiable owner before it reaches an external audience.
Key Takeaways
1. AI Adoption Is a Technology Decision. AI Governance Is a Management Decision.
The two are related, but they are not the same.
2. Professional Responsibility Does Not Change Because Work Is AI-Assisted.
Competence, confidentiality, candor, and supervision remain the governing standards.
3. The Issue Is No Longer Whether AI Can Make Mistakes.
Courts have already demonstrated the consequences when AI-generated outputs are accepted without appropriate review and oversight.
4. Accountability Remains the Defining Question.
The most important issue is not which AI system produced the work. It is who accepted responsibility for it before it reached a client, regulator, or court.
Conclusion
The Future Belongs to Governed AI
The long-term advantage in legal AI will not belong to the firms with the most advanced tools.
It will belong to the firms with the clearest system of accountability.
Technology can accelerate the production of work.
It cannot assume responsibility for its quality.
In the long run, organizations will be judged less by the AI they adopt and more by the governance they establish around it.
A Question for Law Firm Leaders
Every system is ultimately judged by the clarity of its accountability.
If your firm were asked today to explain how AI-assisted work is governed, could it identify a consistent review process and an accountable owner for every client-facing or court-facing deliverable?
Or is governance still assumed to exist simply because the technology does?