Imagine facing a massive iceberg as it rises from the ocean. What we see is only a small part of its size—the largest part, its true mass, is hidden beneath the surface.
This is precisely how two seemingly different forces shaping our world function: artificial intelligence and the legal profession.
Chatbots that instantly answer our questions and the boardrooms where the fates of people and companies are decided represent only the tip of that iceberg.
But what lies beneath the water? What forms the foundation of these complex systems we increasingly depend on?
This article is a journey into the depths—we will attempt to dive below the surface and illuminate the hidden layers that enable this "magic."
We will discover that AI and the law share astonishing similarities in their structure, challenges, and even ethical dilemmas.
Both rely on vast amounts of invisible human labor, complex infrastructure, and systems that amplify biases from the past. Both promise order and predictability in a chaotic world, creating an illusion of absolute certainty.
Through the iceberg analogy, we will see how a "simple" AI response or a "final" legal document actually relies on a web of hidden processes.
Understanding this submerged foundation is crucial not only for using these tools but also for critically evaluating and shaping them as a society.
Let us explore together how these two icebergs float side-by-side, and sometimes even merge, shaping our future.
The Visible Tip: Polished Outputs and User Interfaces
For the average user, AI is accessible and straightforward. We use consumer apps, chatbots like ChatGPT and Bard, voice assistants like Siri, and recommendation engines on Netflix or Spotify. We input questions, voice commands, or images, and we receive answers, playlists, or generated art. The interface is designed to be intuitive, creating an illusion of effortless, objective intelligence.
Similarly, the public face of corporate law is one of polished finality. We see headlines about multi-billion dollar acquisitions, IPOs, and mergers. We see executives shaking hands at closing ceremonies and read about high-profile advisors in the “Wall Street Journal”. The signed contract or the court ruling is the tangible output, the "product" that the client and the public see.
This visible layer is an “illusion of certainty “. It suggests a clean, direct process leading to an inevitable conclusion.
Below the Waterline: The Hidden Machinery
Beneath the simple user interface of AI lies an immense, costly, and ethically complex infrastructure:
Architecture and Training: Complex neural networks with billions of parameters are trained for months on supercomputers, costing millions of dollars. This process consumes energy comparable to small towns.
Data Processing: Petabytes of text, images, and video are scraped from the internet, often without consent, then filtered, labeled, and processed by a global, often low-paid, workforce.
Hidden Human Labor: Behind every AI response lies the work of hidden moderators and data labelers, as well as "trainers" who employ techniques such as Reinforcement Learning from Human Feedback (RLHF) to mitigate bias and toxicity.
Ethical and Operational Burdens: AI systems can amplify societal biases inherent in their training data, act as "black boxes" whose decisions are unexplainable, and remain vulnerable to security attacks, such as prompt injection.
In parallel, beneath the signed legal document lies 90% of the corporate legal work, which remains deliberately opaque:
The Grind of Diligence and Drafting: A "simple" $50 million acquisition rests on 400 hours of due diligence, 68 draft versions of a contract, and young associates working until 3 a.m. in "virtual data rooms."
Coordination of Chaos: Lawyers must manage 15 teams of various specialties across multiple time zones, navigating tax, antitrust, and regulatory issues.
Risk Management and "Weasel Clauses”: The real skill involves designing hidden protection mechanisms: indemnity escrows, warranty insurance, and "hellish" clauses buried in annexes. Lawyers are not deal-makers but risk engineers, deactivating landmines no one will ever see.
The Human Toll: A culture of burnout is normalized, with associates working 2500+ hours per year, where "up or out" promotion structures are standard.
The shared truth is that the polished, visible tip floats entirely because the massive, hidden base absorbs the chaos.
Shared Realities: A Common Iceberg
When we place AI and the legal profession side by side, their parallels become undeniable. They are not just similar in structure; they share fundamental, and often unsettling, systemic truths.
Both fields are built upon a foundational “myth that clashes with reality”. In AI, the myth of "Objective Outputs" obscures a reality of carefully controlled transparency, where settings like "Creative" versus "Precise" modes shape the answers. Similarly, in law, the myth of the "Letter of the Law" gives way to a practice of controlled disclosure and strategic ambiguity, where language is crafted to reveal or conceal.
This polished surface is supported by a vast amount of “often overlooked human labor”. The AI ecosystem relies on operators, often paid as little as $2 per hour, who tag toxic content, and "trainers" who attempt to instill ethical principles through feedback systems. Mirroring this, the legal profession depends on associates working 80-hour weeks, painstakingly reviewing documents, and performing the grueling work of due diligence.
Furthermore, both systems are plagued by “deep-rooted systemic biases”. AI models are trained on the internet's vast corpus, which includes sexist and racist content, leading them to perpetuate and amplify these stereotypes. The law, in turn, operates on a foundation of historical precedents from discriminatory eras, where "neutral" clauses can have a disproportionately negative impact on marginalized groups. For both, the ideal of true neutrality is a myth, as they inevitably embed historical injustices into their modern operations.
These structures also share “critical systemic weaknesses”. AI's progress is vulnerable to physical constraints like shortages of advanced computer chips (GPUs) and a massive dependency on the energy grid. The legal system faces its own forms of paralysis through court backlogs, a lack of judicial staff, and the tactical abuse of procedural rules like discovery.
“Ethical gray zones” are another common frontier. In the AI world, this manifests as deepfakes being defended under the guise of free speech and legal loopholes being exploited as a competitive weapon. In the legal profession, similar gray zones include creating tax havens that are technically "compliant" with local laws or drafting agreements that provide compensation for polluters.
Finally, at their core, both are complex systems whose primary function is “risk management through “black boxes”. AI uses complex heuristics to curb its own "hallucinations" and is specifically trained to handle rare, edge-case scenarios. Likewise, legal contracts are built with protective mechanisms like "Material Adverse Effect" clauses and hidden triggers in “force majeure” sections, designed to manage unforeseen crises.
Shared Truths:
- Solutions rely on underpaid, labor-intensive engagement.
- "Neutrality" is a myth—both systems embed historical injustices.
- Outcomes depend on fragile systems that can be manipulated.
- Avoiding catastrophe is more valuable than creating visible value.
Conclusion: A Foundation for Responsible Innovation
The iceberg analogy does not condemn either AI or the law. Instead, it reveals their shared reality as complex systems built by humans. Both rest on intricate foundations—hidden technical and legal work that enables refined outputs. Both are built upon existing systems (datasets/legal precedents) that inform modern decisions. Both contain an inevitable human element, from the subjectivity in training data and legal interpretation to the risk-management decisions made for unforeseen circumstances.
Furthermore, they are interdependent. AI relies on legal frameworks for intellectual property and liability, while the legal profession increasingly adopts AI tools for document review and case prediction.
Their combined future is not a warning—it is an opportunity:
For AI: To adopt the rigor of the law in governance, ethics, and accountability.
For the Law: To embrace the efficiency of AI while preserving essential human judgment.
The shared iceberg is not just a hidden cost; it is the essential foundation for responsible innovation. By understanding and strengthening these submerged foundations, we can ensure that the visible tips of both AI and law are not just illusions, but reliable and trustworthy pillars of our society.
By Nemanja Stepanovic, Managing Director, JPM & Partners
