Crosby wants malpractice insurance for its AI lawyers, not just its human ones

Lawyers carry malpractice insurance as a matter of routine. Crosby, a New York-based AI-native law firm, now wants its software to carry it too.

The firm announced in late July 2026 that it plans to secure professional liability insurance for its artificial intelligence agents. That puts it on a path toward legal work that needs far less human supervision than the industry has been willing to accept so far.

Founder and CEO Ryan Daniels made the announcement on LinkedIn. The plan is to insure the AI agents almost as if they were autonomous professionals, rather than treating them as tools a lawyer happens to use.

How Crosby works today

For now, humans still sit at the end of every assembly line. Crosby employs somewhere between 45 and 60 licensed attorneys, and they review every legal output the AI produces.

That output is mostly contracts. Think non-disclosure agreements and master services agreements, the paperwork every growing company signs by the stack and nobody enjoys reading.

The AI agents handle the bulk of the contract review. The lawyers act as quality control, catching what the machines miss and signing off on the finished product.

Crosby’s structure is what makes the insurance question possible in the first place. The company runs as two entities: a technology company and a licensed professional limited liability company, or PLLC.

Put simply, Crosby is a real law firm, not a software vendor selling a contract-review app. It already holds malpractice coverage and delivers services through a hybrid of code and conventional legal practice.

The pitch has found buyers. Clients include AI startups such as Cursor and Runway.

The growth numbers are steep. Crosby’s contract volume climbed from $30 million to $1 billion in under a year, reaching that mark by mid-2026.

Investors noticed earlier. The firm raised a $60 million Series B in the spring of 2026, at a valuation of roughly $400 million.

Why insuring an AI agent is harder than it sounds

Malpractice insurance is built around a simple premise. A licensed professional makes a judgment, and if that judgment is negligent, the policy pays.

An AI agent complicates every part of that sentence. It holds no license, it is not a person, and its judgment is the product of training data and model weights rather than law school and bar exams.

So Crosby cannot simply call a broker and ask for a quote. The firm says it is working with auditors, bar associations, state regulators and insurers to build new evaluation standards and policy terms from scratch.

The goal, as the firm frames it, is “insured outputs” that its AI agents can produce on their own. In other words, a contract reviewed by software that comes with the same financial backstop as one reviewed by a partner.

The broader legal tech picture

Crosby is not the first company to explore insurance in this corner of legal tech. Earlier efforts from Orbital and Corgi touched on similar territory.

The difference is in the framing. Crosby’s approach treats its AI agents as close to autonomous entities, rather than as instruments that sit firmly under a lawyer’s control.

Crosby has also been building toward this for a while. The firm has a track record of prior work on insurance that predates this announcement.

What this means

The most immediate stakes are for law firms and legal tech startups watching from the sidelines. If Crosby can get insurers and regulators to agree on how to evaluate and price AI agent risk, it creates a template others could copy.

The economics are just as important. Crosby’s model depends on a relatively small bench of attorneys overseeing a growing volume of work. If insured AI outputs reduce how much human review each contract needs, the firm could scale volume faster than it scales headcount.

There are real risks, though. Bar associations and state regulators have not yet signed off on standards for insuring AI agents, and the process Crosby describes involves multiple parties with very different priorities.

Insurers will want data on error rates and audit methods before committing to policy terms. Regulators will want to know who is accountable when an insured AI agent gets something wrong, and whether a client fully understands that no human may have reviewed the work.

What to watch next is which regulators and insurers publicly engage, and what evaluation standards emerge from the talks. The terms of any eventual policy, including what it covers and what it excludes, will say a lot about how much autonomy the industry is actually prepared to grant AI agents.