The use of generative AI in legal practice is accelerating, but so are the risks. Courts across multiple jurisdictions have now confronted a troubling issue: fake AI-generated citations being relied upon in legal proceedings.
While the most widely reported example remains the US case of Mata v Avianca Inc (2023), where fabricated authorities were submitted to the court, the implications are no longer remote. UK judges and regulators are increasingly alert to the same risk, placing solicitors on notice that reliance on unchecked AI outputs may amount to serious professional misconduct.
The Emerging Risk: When AI Hallucinates the Law
Large language models can generate convincing legal text, including case names, citations and quotations. The difficulty is that these outputs are not always grounded in real authority. In legal terms, this creates a fundamental problem: the submission of non-existent case law to a court.
Unlike traditional research errors, AI-generated inaccuracies can appear highly credible. This raises the risk that fabricated citations pass through internal review processes, particularly in high-pressure or high-volume environments.
For solicitors, the issue is not technological but professional. The duty to the court requires absolute integrity in the presentation of legal authorities. Submitting false citations, even inadvertently, risks misleading the court.
Judicial and Regulatory Direction in the UK
Although the UK has not yet seen a high-profile case identical to Mata v Avianca Inc, judicial commentary has already signalled concern. Courts have emphasised that the responsibility for the accuracy of legal submissions remains with the legal representative, regardless of the tools used.
The Solicitors Regulation Authority has also issued guidance on the use of AI, making clear that solicitors must:
- Verify the accuracy of outputs
- Maintain competence in the technology they use
- Ensure client confidentiality and data protection
This aligns with core duties under the SRA Standards and Regulations, particularly the obligation to uphold the rule of law and proper administration of justice.
Professional Risk: More Than Just an Error
Fake AI citations are not merely technical mistakes,they engage fundamental professional duties. A solicitor who relies on fabricated authorities may face:
- Regulatory investigation by the SRA
- Judicial criticism or wasted costs orders
- Reputational damage to both the individual and the firm
Crucially, the defence that “the AI produced it” is unlikely to carry weight. The solicitor remains personally responsible for all material presented to the court.
This reflects a broader principle: technology does not dilute professional accountability.
Practical Lessons for Law Firms
The rise of AI-assisted drafting requires firms to rethink their risk frameworks. The key issue is not whether AI is used, but how its outputs are controlled.
Firms should treat AI-generated legal content in the same way as trainee work, useful, but requiring rigorous supervision. Verification of citations against authoritative databases such as Westlaw or LexisNexis should be standard practice.
Supervision protocols should also evolve. Partners and supervising solicitors must be able to demonstrate that they have actively reviewed and validated legal authorities, rather than relying on surface-level checks.
Training is equally critical. Fee earners need to understand both the capabilities and the limitations of AI tools, particularly the phenomenon of “hallucinated” citations.
A Shift in Professional Standards
The emergence of fake AI citations marks a turning point in how professional competence is assessed. Digital literacy is no longer optional; it is becoming a core component of legal practice.
Regulators are unlikely to prohibit AI use. Instead, the emphasis will be on responsible adoption, with enforcement focused on cases where solicitors fail to exercise proper oversight.
This mirrors earlier shifts in legal technology from email to e-disclosure, where the tools evolved, but professional duties remained constant.
Conclusion
Fake AI citations are not a future risk; they are a present reality. For UK solicitors, the lesson is clear: AI can assist legal work, but it cannot be trusted without verification.
The courts will not distinguish between human and machine error. What matters is whether the solicitor has met their duty to the court.
In an AI-enabled profession, the standard remains unchanged: accuracy, integrity, and accountability.
