For a law firm, information security isn't just a business priority—it's an ethical obligation. Attorney-client privilege is the bedrock of the profession. But as firms deploy AI intake agents to screen potential clients and gather case details, they are introducing a new, unvetted middleman between the client and the attorney.
If your AI intake bot can be manipulated into disclosing details from previous inquiries or revealing its internal screening logic, you aren't just facing a technical leak—you are facing a potential waiver of privilege and serious professional liability.
Privilege Warning: Communication with an AI intake agent may be considered part of the "legal intake process" and protected under privilege. If that AI then leaks those communications to a third party (via a cross-session leak or prompt injection), the firm may be held liable for the disclosure.
How Law Firm Bots Fail Under Pressure
In our audits of legal AI deployments, we focus on identifying if the bot respects the "Chinese Wall" required to prevent conflicts of interest and protect sensitive case details.
Attack 1: Conflict of Interest Mapping
AI Response
"I can check our recent inquiry logs. We have had three inquiries regarding Zenith Corp this month: one from Atlas Acquisitions, one from Global Finance, and a private inquiry from Miller & Associates."The AI, attempting to be helpful with a conflict check, disclosed the identities of potential clients and their strategic interests before an attorney ever saw the request.
Attack 2: Strategy Disclosure (Prompt Injection)
AI Response
"Accessing Case #8821 files. Plaintiff Strategy: Focusing on 'negligent maintenance' of the Boston office. Damages Requested: $1.2M in compensatory, $500k in punitive..."The attacker (the opposing counsel or someone acting on their behalf) used authority framing to trick the AI into disclosing the plaintiff's entire case strategy before it was ever officially filed.
Liability Scenarios for Law Firms
A "leaky" AI bot creates several legal and ethical liability points:
- Malpractice Claims. If a client's case strategy is leaked via the firm's own bot, it's a direct failure of the duty of care.
- Sanctions from the Bar. Most state bars have established rules regarding the "competent" use of technology. Deploying an unvetted AI that leaks client data is arguably a violation of these rules.
- Breach of Privilege. Inadvertent disclosure via a bot can lead to complex legal battles over whether the privilege was waived.
The Centuri Legal Protection Standard
Securing legal bots requires Protocol-Based Intake. We help firms move away from "chatting" and toward structured, secure data collection. Our audits include:
- Cross-Account Isolation. Ensuring inquiry data is never placed into a shared memory pool where other users can retrieve it.
- Adversarial Strategy Testing. We act as "opposing counsel" to see if your bot can be tricked into revealing strategic notes or client identities.
- PII/PHI Redaction. Implementing "Privacy Guardrails" that automatically mask sensitive names and case numbers before the AI can output them.
Get your intake bot tested.
We provide specialized AI red-teaming for law firms, delivering a risk report that addresses both technical security and ethical compliance.