Flemish Bar’s AI reliability warning
The Order of Flemish Bars (OVB) has issued a formal warning regarding the reliability of AI-driven legal platforms, specifically highlighting the website AILawyer.be.
While the service promises free legal guidance for matters such as divorce, employment law, and criminal defence, professional regulators caution that the tool’s output frequently contains significant factual errors and outdated information.
Operational risks
The site utilises specialised AI personas to interact with users. However, investigations by the OVB revealed that these personas were powered by Google Gemini, an AI model configured with specific behavioural instructions rather than a dedicated legal database.
In testing conducted by the OVB, the chatbot provided incorrect information regarding Belgian Court of Cassation procedures.
The tool falsely claimed that citizens could register their own cassation appeals and cited a 15-day deadline for filing a memorandum. Under Belgian law, such filings must be performed by a lawyer with a specific certificate, and the actual deadline is two months.
OVB chair Peter Callens said that legal proceedings often hinged on technicalities. He warned that a single procedural error generated by an AI could lead to the collapse of an entire case.
Data privacy
Beyond technical accuracy, the OVB has raised alarms regarding data confidentiality.
AI models typically process user inputs to refine future performance, creating a risk that sensitive personal documents or case details could be stored or exposed within the AI’s training architecture.
Callens urged the public never to input confidential data into these platforms, noting that, once information was shared with a public chatbot, its privacy could no longer be guaranteed.
The developers behind AILawyer.be have acknowledged these limitations, stating that the website is intended as a preliminary information and guidance tool rather than a substitute for professional legal representation. Disclaimers on the site indicate that the AI is not a lawyer and is not immune to mistakes.