AI in courts and business: is the revolution ending in convictions?

The use of artificial intelligence carries serious legal risks, as evidenced by the July 2026 verdict against the CEO of an AI start-up for insider trading and the growing warnings from Polish authorities regarding abuses of the Grok model. This technology has ceased to be a safe haven for innovation, becoming an area where every algorithm decision can lead to criminal liability. Companies that have not implemented rigorous data control mechanisms are now facing the real threat of management losing their freedom.

The evolution of the courts' approach to technological innovation does not resemble a calm process of adaptation. It is more of a violent collision with a reality where code becomes evidence in a criminal case. On July 9, 2026, the tech world received a signal that cannot be ignored. The conviction of the CEO of a start-up specializing in AI solutions for insider trading is not just an isolated case. It is the definitive end of the period in which algorithm creators could hide behind the "incomprehensibility" of the machine learning process. The court ruled that exploiting advantages resulting from access to data that should not be publicly available is a classic stock market crime, regardless of whether the tool used for this purpose had "AI" in its name.

This verdict puts an end to the "black box" narrative, for which no one is held responsible. In practice, this means that managers must stop treating AI models as autonomous entities making decisions in a vacuum. If an investment algorithm gains access to information that would normally be protected by a confidentiality clause, the responsibility for this breach lies directly with the company's board. Start-ups that scaled their solutions without ethical oversight must now prepare for audits they previously did not consider.

The Polish Ministry of Digital Affairs recognized the threats stemming from automation much earlier than the level of public debate would suggest. As early as January 5, 2026, the Minister of Digital Affairs called for the immediate signing of regulations governing online content, pointing directly to the risks associated with abuses of the Grok model. The problem is not just disinformation. It is about the fact that generative tools have become a sophisticated instrument for bypassing legal safeguards that were intended to protect the information space.

Authorities in Warsaw are reacting to a phenomenon where technology is beginning to outpace existing civil frameworks. The current Polish Civil Code, created in times when no one thought about algorithms deciding on social life, puts prosecutors against a wall. How to prove guilt when authorship is blurred between programmers, users, and the learning system itself? The Ministry is pushing for changes, but lawyers point to a deep conflict between the letter of the law and the technological flexibility of the Grok model. Every generative tool has its "tipping point," where it stops providing knowledge and starts manipulating facts. The Ministry's warnings are therefore not just a technical appeal for a new law. It is a signal that the state is beginning to view AI systems not as optimization tools, but as a potential threat to public order.

Trust in technology in Poland rests on fragile foundations. On one hand, we have ambitions to be an innovation hub, and on the other – a judicial reality that increasingly treats "digital footprints" as key evidence in cases of fraud or personal rights violations. If the law is not quickly tightened, courts will be forced to create precedents on their own, which for business means a lack of predictability. And in the corporate world, a lack of predictability means higher legal costs and an outflow of investors.

The conflict between privacy and state security takes on a new dimension when we include smartphones and the data contained within them in the equation. The Tommy Robinson case from October 13, 2025, in which he refused to provide the police with his phone PIN, became the starting point for a broader debate about the limits of law enforcement interference. Robinson argued that locking the device protects journalistic materials. This approach forces the justice system to answer the question: where does the right to privacy end in the era of total digitalization?

For a lawyer dealing with new technologies, the Robinson case is a lesson that a phone is no longer just a communication tool, but a "black box" of private and professional life. Services increasingly want access to it, arguing this is for fighting crime, while citizens defend themselves using arguments about professional secrecy. This tension is compounded by the fact that these devices have AI models installed that can analyze, organize, and draw conclusions from data that the phone owner might not even have known existed.

The risk is not abstract. If we combine this with the July verdict for the AI start-up CEO, the picture becomes clear: a digital footprint is the hardest evidence a prosecutor can possess. In cases where AI acts as an "accomplice" in breaking the law, judges will not ask about the programmer's intentions. They will look at the effects generated by the system. This is a signal for every business lawyer: in abuse cases, a defense line based on "algorithm error" is already ineffective.

European legal standards, shaped by rulings from Luxembourg, are increasingly affecting the Polish reality. The ruling of the EU court advisor from April 4, 2025, ordering Poland to recognize same-sex marriages concluded abroad, shows that European pressure on the Polish legal system goes beyond technological issues. However, this is part of the same process: forcing Polish law to be open to external standards, which is particularly noticeable in the age of AI.

Lawyers with many years of experience, analyzing the evolution of law in the face of digital challenges, often refer to studies such as the review in "Przegląd Sądowy" from 2021/2022. Even then, the inevitability of the evolution of law in the face of technology that does not recognize national borders was pointed out. Today, these predictions have become everyday life. Polish legislation, instead of anticipating trends, is desperately trying to keep up with rulings that have already been made. This is not an evolution of the system, but putting out fires caused by algorithms that do not ask the legislator for permission to operate. If the market does not understand that criminal liability is written into the code, subsequent verdicts will only confirm that the law, in a clash with AI, knows no leniency.

The question is: is business able to manage this risk? The answer is: yes, but it requires a paradigm shift in the approach to implementing AI. Instead of counting on "transparent rules" and good intentions, managers must introduce hard control tools.

My advice for a manager: implementing an AL-Compliance (Algorithmic Liability Compliance) audit is essential. Such an audit must cover three pillars. First: full mapping of the data sources on which the model is trained — you must know if you are not using confidential competitor information or sensitive data. Second: introducing a "human-in-the-loop" procedure for every decision generated by AI that has financial or legal consequences. The algorithm cannot be the sole decision-maker. Third: establishing a "kill-switch" protocol that allows for immediate cutting off of the model's access to external data the moment irregularities in its operation are detected.

For a lawyer: treat AI not as a tool, but as an employee for whom you bear full civil and criminal liability. In contracts with AI system providers, you must force clauses on full indemnification for system errors and require access to model operation logs so that, in the event of an inspection, you can demonstrate that the company exercised due diligence in supervising the machine. The era of thoughtless use of "magical" AI solutions is over. Today, in court, the winner is not the one who has a better model, but the one who can prove that they are in control of it.

Questions and answers

Is AI safe support for lawyers?

It requires enormous caution. The case of the AI start-up CEO from July 2026 clearly shows that technology can be used for illegal activities, and the responsibility for these acts falls on the people managing the tool, not the algorithm itself.

What is the scandal related to the Grok model?

These are concerns regarding abuses in content generation, which prompted the Polish Minister of Digital Affairs to accelerate work on online content law as early as January 2026, due to the risk of manipulation and bypassing legal safeguards.

Is refusing to provide a phone PIN effective in court?

It is a point of contention. The Tommy Robinson case from October 2025, who argued his refusal was to protect journalistic materials, shows that the line between the right to privacy and the requirements of law enforcement remains a field of sharp legal conflict.

What challenges does AI pose for civil law?

The main problem is adapting the concept of guilt. The current system is based on human responsibility for words, while in the case of AI, the source of the generated content often remains elusive, which forces courts to patch regulations on an ad-hoc basis with case law.

Does European case law influence Polish law on AI issues?

Yes, rulings from the EU, such as the one from April 2025, force the Polish legal system to redefine concepts and adapt to European standards, which, combined with technological challenges, creates pressure for rapid legislation.