I just got married a month ago.

Our honeymoon was an investment banking conference in Puerto Rico. That is what happens when you marry someone who fully accepts that "vacation" may involve panel discussions, investor dinners, and at least one conversation about governance structures before noon.

It ended up being a genuinely interesting week. I got the chance to speak with a lot of smart founders working on technology and projects that are going to change the world. And maybe because I had been thinking a lot about commitment recently, one pattern kept coming up in nearly every conversation.

A lot of founders have replaced the 2 AM call to outside counsel with a prompt to Claude or ChatGPT.

That is not even the worst version of the problem.

The Worst Version

I talked to a founder last week who used AI to draft their customer MSA. The AI gave technically accurate language. It just did not know their sales team had been verbally promising something completely different for eight months. The contract was fine. The liability was not.

That story is not unusual. But there is a second problem underneath it that I do not think most founders have fully processed yet, and earlier this year a federal judge made it very real.

The Ruling That Should Be on Every Founder's Radar

In February 2026, Judge Jed Rakoff of the Southern District of New York issued a ruling involving a former CEO named Bradley Heppner. Heppner was facing a federal securities fraud investigation and, like a lot of smart executives trying to think through complicated situations, he turned to AI for help. He used Claude to research legal questions connected to the investigation, generated dozens of pages of prompts and responses, and eventually shared those materials with his lawyers at Quinn Emanuel.

At some point, I think a lot of people reading this probably assume the story goes:

"Well, once the lawyers had it, it became privileged."

It did not.

The government obtained the AI conversations. Defense counsel argued privilege applied. Judge Rakoff rejected the argument completely.

And the reasoning was remarkably straightforward.

Claude was not acting as a lawyer. It was not a licensed legal adviser. It did not owe fiduciary duties. It was not bound by professional confidentiality obligations. And maybe most importantly, Heppner had voluntarily shared sensitive information with a third-party commercial platform whose own terms made clear that prompts could be collected, stored, and disclosed.

The court looked at the situation and essentially said: you typed your legal strategy into a commercial technology platform. That is not the same thing as speaking confidentially with counsel.

What makes the ruling unsettling is not just the privilege issue itself. It is how normal the behavior feels now. Because this is exactly how people are starting to interact with AI. Not as software. As thought partner.

People brainstorm with it. Confess uncertainty to it. Test theories with it. Work through sensitive situations with it in the middle of the night when nobody else is awake and the pressure feels heavy.

And the more human these systems start to feel conversationally, the easier it becomes to forget that legally, they are still platforms. Not advisers. Not fiduciaries. Not protected relationships.

That distinction matters a lot more than most companies realize yet.

The ruling is not limited to criminal law. The same logic spills into internal investigations, employment disputes, compliance reviews, civil litigation, regulatory inquiries, privacy analysis, and governance discussions.

A compliance employee trying to work through a potential GDPR issue with ChatGPT. A founder testing antitrust questions through Claude before a board meeting. An employee asking AI whether their compensation agreement is enforceable before forwarding it to HR.

All of those interactions may create discoverable material. And forwarding the output to your lawyer afterward does not magically convert it into privilege any more than forwarding your Google search history would.

The strange thing is that most founders are not using AI this way because they are reckless.

They are using it because it feels efficient, available, intelligent, and private.

But those are not the same thing.

Context and Judgment Is the Whole Job

The AI discoverability problem is a symptom of a larger one. Both AI tools and transactional outside counsel share the same core limitation: they only know what you tell them.

They do not know your company the way an embedded legal team does.

They do not know that the deal everyone is excited about has the same energy as the one that created a mess two years ago. They do not know your Head of Sales has been describing the product differently than your Product team for the last six months. They do not know the board has quietly become nervous about governance issues nobody wants to address directly yet. They do not know which risks theoretical and which ones are are already starting to form patterns.

Good legal advice is rarely just about the document sitting in front of you.

Most of the time it is about context. It is about knowing that Sales has been promising something Product never fully approved. It is about knowing the board is already nervous about a certain issue before the issue formally appears. It is about recognizing that the same liability provision keeps appearing in customer contracts because somebody internally is creating pressure that has not yet been openly discussed. It’s about knowing that Rich in Procurement is smart but a jerk and we don’t ever want him on the stand.

Companies are strange little ecosystems. They have personalities. They have habits. They have old wounds. They have historical decisions that made perfect sense three years ago and make absolutely no sense anymore, except nobody wants to revisit them because the company is moving too fast and everyone is tired.

The best legal relationships are not built around emergencies. They are built around proximity. Around pattern recognition. Around having someone close enough to the business to notice what is quietly developing before it becomes a real problem.

This Is Especially True Right Now

Most AI companies and growing technology businesses are moving much faster than their governance structures, privacy frameworks, or internal approval systems were originally designed to handle.

Marketing is talking about AI one way. Product is describing it another way. Investors are asking increasingly sophisticated diligence questions. States are passing new laws almost monthly. The FTC is paying attention to claims around AI capabilities and automation.

Meanwhile, most founders are still managing legal reactively because that is how they managed it when the company was smaller.

But at a certain point, legal stops being primarily about documents and starts becoming operational. It becomes part of how decisions are made. Part of how risk is evaluated. Part of how leadership thinks through growth, hiring, partnerships, product development, and governance.

Founders Want Two Completely Opposite Things at the Same Time

Speed, but also protection.

Flexibility, but also predictability.

Legal involved, but Legal out of the way.

A Ferrari at used Volkswagen prices.

I get it. Nobody wakes up excited to spend money on lawyers. Especially in the early stages when every dollar matters and legal feels like something you only need when things go wrong.

But eventually every growing company hits the point where legal stops being a cost center you reluctantly call and starts becoming infrastructure.

If legal only appears when something is already on fire, then legal becomes emotionally associated with friction, cleanup, delay, and bad news. When legal is close enough to the business to participate before the explosion, something different happens. Problems get spotted earlier. Conversations get handled differently. Small issues stay small.

The best in-house lawyers are not just technically good lawyers. They are people who have seen enough versions of the movie to recognize certain scenes before everyone else realizes what kind of movie they are in.

Why We Rebranded

That realization is part of why we recently rebranded from Unified Law to Unified CLO.

For years, we kept finding ourselves explaining what we were not. We were not a traditional law firm. We were not a referral network. We were not outside counsel people called when something already went wrong.

The word "law" carried baggage. It implied a service you called. A vendor. A transaction. An invoice attached to a problem.

But that was never how we worked.

The companies we work best with do not treat legal like a detached service sitting outside the business. They treat it as part of the business itself, part of operations, strategy, growth, governance, hiring, product, risk, and decision-making.

That is what a Chief Legal Office is.

Not something you call once the fire starts. Something you already have before the smoke shows up.

A month ago I made a commitment to a person I trust completely with the hard stuff, the long view, and the decisions that matter. He knew what he was signing up for when he agreed to a honeymoon at an investment banking conference.

The founders reading this know what they are signing up for when they keep treating legal like a 2 AM text.

Both are a choice. Only one ends well.

One shows up when something goes wrong. The other is built to stay.