Who Is Liable for Autonomous AI Agents?
Somewhere on a corporate server right now, an AI agent is negotiating a contract, executing a payment, or making a hiring decision, without a single human clicking approve.
This is no longer a hypothetical. As of June 2026, autonomous AI agents are operating on behalf of companies across global supply chains, financial systems, and employment processes. And when something goes wrong — a hallucinated contract, a rogue payment, a discriminatory hiring pattern — the legal question is immediate and unavoidable: who is responsible?
The answer is both simple and alarming. The AI cannot be sued. It has no legal identity in any jurisdiction on Earth. Every legal consequence of its autonomous actions falls directly on the company that deployed it.
In this deep dive, TRUST. Attorneys at Law unpacks the defining legal tension of the AI age: 21st-century autonomous agents colliding head-first with legal frameworks built exclusively for human beings.
What this episode covers:
We start with the foundational question of legal identity — the critical distinction between a tool, which processes information when directed, and an AI agent, which perceives, reasons, and acts autonomously toward long-term goals. That distinction changes the entire mechanism of liability.
From there, we move into contract law and the doctrine of apparent authority — why companies can be legally bound to deals their AI hallucinated, and why claiming “the computer did it” fails as a legal defence in both common law and civil law systems.
We examine the Knight Capital collapse of 2012 as a precursor: a software glitch that cost $440 million in 45 minutes. Now multiply that across global supply chains with AI agents executing thousands of contracts before a human returns from lunch.
The episode addresses autonomous payments and the specific danger of crypto assets — where blockchain immutability means a mistaken transfer is mathematically irreversible, with no chargeback mechanism and no central authority to call.
We map the three diverging global regulatory philosophies: the EU’s prescriptive risk-based model under the AI Act and the new Product Liability Directive, the Anglo-American sectoral ex post approach, and the Asian state-led governance model — and what compliance looks like when your operations touch all three.
Finally, we reach the frontiers the law has not yet solved: algorithmic cartels formed without human intent, the intellectual property blindspot that vaporises AI-generated assets the moment they are created, criminal exposure for executives who deploy reckless systems, and the unresolved insolvency question — what happens when a bankrupt company’s AI agents keep running, committing to thousands of new contracts for a corporate master that no longer exists.
The law hasn’t changed. It still demands strict human accountability. What has changed is the speed and autonomy at which that accountability is tested.
This deep dive was produced by TRUST. Attorneys at Law. To understand how these frameworks apply to your business, contact us at www.trust.fi.
Jan Lindberg, Partner