Navigating AI in Professional Services: Adnovate’s Confidentiality-First Approach

Artificial intelligence offers the potential for tremendous productivity gains for consulting firms, but must be used cautiously and responsibly. At Adnovate, we have developed a clear policy to harness AI’s benefits while protecting client confidentiality, a concern that’s increasingly moving from theoretical risk to practical reality.

The Core Risk

Whilst there is currently no publicized precedent in biopharma consulting for litigation related to breach of client confidentiality as a result of using public AI platforms, lessons can be learned from other sectors where confidentiality is paramount. Legal regulators and legal-ethics bodies across the UK, Australia, and California have explicitly warned that uploading client confidential information into public AI systems can constitute a breach of confidentiality or privilege waiver [Goraya et al, 2025, Harkness, 2025]. Critically, liability rests with the professional firm, not the AI vendor. Legal precedent treats disclosure to public AI platforms similarly to disclosure to any uninstructed third party, destroying privilege and triggering potential exposure under NDA breaches, negligence, breach of contract, or data-protection violations. While most disputes remain at the advisory stage today, data privacy claims against service providers are already being brought when tech use leads to confidential data exposure [Fisher Phillips, 2025].

Where AI Can Be Used Safely and Where It Cannot

At Adnovate, we distinguish between safe and unsafe use of public AI platforms. Public domain information including scientific literature, published trial results, established mechanisms of action, and regulatory frameworks can be processed safely through public AI systems. This supports background research, client and consultant onboarding, and technical knowledge synthesis.

Confidential information is off limits. Meeting minutes, client calls, regulatory interactions, asset-specific data, clinical development plans, and slide decks containing proprietary details cannot be uploaded to public AI systems, regardless of anonymization, without explicit client consent.

Adnovate’s Policy in Practice

Adnovate actively solicits client preferences regarding AI and LLM use at the start of each engagement. We have already encountered clients who explicitly prohibit AI-assisted meeting minutes due to business sensitivity, a clear signal that confidentiality concerns are real and warrant proactive safeguards.

Our formal position is straightforward: we encourage innovative AI use for non-confidential project tasks but strictly prohibit processing of client- or asset-confidential materials without explicit authorization. When in doubt, our teams escalate to project leadership whose default position is always conservative.

Looking Forward

As regulatory scrutiny intensifies, AI policies are likely to become standard audit requirements. By establishing clear guardrails now, Adnovate demonstrates diligence to clients and protects both parties from emerging legal risks.

The principle is simple: assume all information uploaded to a public AI platform becomes permanently public. If you wouldn’t leave it on a train or discuss it in a restaurant, don’t upload it.

References:

Fisher Phillips (2025. AI Call-Monitoring Lawsuits Are Heating Up: 5 Steps Your Business Can Take to Minimize Risk. Available at: AI Call-Monitoring Lawsuits Are Heating Up: 5 Steps Your Business Can Take to Minimize Risk | Fisher Phillips

Harkess, J (2025). The AI Ethics Landscape: A Map for Legal Professionals. LinkedIn Pulse. Available at: https://www.linkedin.com/pulse/ai-ethics-landscape-map-legal-professionals-dr-jason-harkess-0gstc

Goraya, A, Makhani, A-K, Niganm, S (2025). Digital strangers in litigation: does sharing with AI breach privilege?. International Bar Association, 2025. Available at: https://www.ibanet.org/Digital-strangers-in-litigation