All essaysEssay - 18 February 2026

Lobsters: the autonomous overnight agents reshaping legal work

Last month, an Austrian developer named Peter Steinberger published a side project.

By Quentin Solt - about 3 minutes to read

Last month, an Austrian developer named Peter Steinberger published a side project. It runs on a Mac Mini costing around £500. It works through the night while you sleep. By morning it may have read thousands of pages, drafted documents, organised files and – in one widely shared example – negotiated a $4,200 discount on a car purchase entirely by email, without human intervention.

Within weeks the repository reportedly attracted over 100,000 followers on GitHub. OpenAI hired Steinberger with notable speed. The project is called OpenClaw, and its cheerfully eccentric mascot is a lobster. Hence the headline.

What it actually does

OpenClaw is an autonomous AI agent.

The difference between this and the AI tools many lawyers already use is straightforward: you don't have to be there. You give it instructions, it connects to your emails, calendars, documents and messaging apps, and it acts. It doesn't wait to be prompted. It doesn't stop when you go to bed.

Crucially, no new AI model was involved. The underlying models already existed. Steinberger simply removed the throttle and let them run overnight. That is the structural shift we are living through – and it is emerging faster than almost anyone predicted.

What this means for the economics of legal practice

Law firm economics are built on leverage: senior expertise directing junior time, all billed by the hour. Autonomous agents don't only disrupt at the junior tier. They attack the entire stack simultaneously.

A sole practitioner with a well-configured agent can now produce overnight what previously justified a small team:

  • first-draft agreements

  • due diligence summaries

  • risk registers

  • correspondence logs

  • an overnight share purchase agreement review

  • bulk lease abstraction

  • disclosure bundle triage

The time-rich individual is beginning to outproduce the capital-rich institution as the marginal cost of cognitive labour collapses.

The regulatory position – protected for now, but only for now

Regulated solicitors currently occupy a defensible position. The SRA Code of Conduct requires competence and supervision.

Legal professional privilege attaches to advice given by qualified lawyers acting in that professional capacity.

Professional indemnity insurance provides a safety net that unregulated providers cannot match. But, in reality, most clients assume they will never need to rely on that protection.

Regulators and insurers are not yet asking the right questions about autonomous agents. Lawyers are (wisely) anticipating that they will, and that those questions will be uncomfortable: who supervised the overnight output? Who verified the research before it went to the client? Who approved the draft?

The “AI did it" will not satisfy the SRA, a negligence claimant or a judge.

Solicitors who have deployed these tools carelessly will find that out at the worst possible moment.

The immediate risk – and it isn't the technology

What should concern lawyers more than the technology itself is that clients will be approached by unregulated business advisors, management consultants and AI-native providers offering something that looks like legal work – quickly, cheaply and outside the regulatory framework entirely.

No professional indemnity. No SRA oversight. No privilege. No duty of care enforceable in the way a solicitor's duty is enforceable.

But to a cost-conscious client, willing to take a commercial view, that proposition may be genuinely compelling. Fast, inexpensive and apparently competent. Clients may accept the risk of a flawed clause, an unregistered interest, a privileged communication that turns out not to be privileged at all, in exchange for accelerated advice and reduced cost.

The bottom line

The lobsters have arrived. They are not going back. Their pincers are strong.

Solicitors have long viewed regulation and professional insurance as protection – a mark of quality that distinguishes them from less scrupulous alternatives. They may increasingly find that the same frameworks prove to be a constraint on competing with emergent, unregulated alternatives wielding identical tools at a fraction of the cost.

The window in which the traditional protection still means something to clients is open. It will not remain so indefinitely.