As privacy rules tighten, secure infrastructure becomes the new gatekeeper for attorneys wary of exposing client data to language models.

SAN FRANCISCO: #AICompliance--When partners at a California law firm floated the idea of using generative AI to draft briefs and summarize case law, their IT department blocked the plan. The reason: feeding client data into public AI models could expose privileged information and violate bar rules. The promise of faster research collided with the profession’s deepest taboo: confidentiality.

After months of evaluation, the firm piloted a confidential-compute platform from Phala - one of several emerging infrastructure providers building “trust layers” for AI. The system allows models to process encrypted data inside secure enclaves, where information remains sealed even from the infrastructure operator itself.

“It gave us verifiable proof that no information ever left our control,” said one partner involved in the pilot. “That changed the conversation from can we use AI? to how safely can we deploy it?”

AI Adoption Meets Legal Ethics

Both the American Bar Association’s Formal Opinion 512 (2024) and the California Bar’s Practical Guidance on Generative AI (2023) reaffirm that attorneys remain responsible for maintaining confidentiality, competence, and supervision when using third-party AI systems. These requirements have left many firms unable to leverage modern tools that rely on external data processing.

Confidential computing offers a new compliance path. By isolating AI models in secure enclaves and generating cryptographic attestation-verifiable proof of where and how data is processed-firms can demonstrate control over sensitive information during every stage of an AI workflow.

From Experiment to Standard Practice

In the California pilot, the firm deployed Phala’s confidential AI for document review and case analysis. Attorneys could spin up private model instances by the hour, process discovery materials securely, and generate summaries in minutes. All data remained encrypted end-to-end, and the firm received automated attestation reports for audit purposes.

The results:

• 100% compliance with state bar confidentiality rules
• 40% faster document review
• Predictable costs through hourly, isolated model use

“The legal profession is realizing that privacy-by-design infrastructure is not just a cybersecurity measure-it’s a business enabler,” said Marvin Tong, CEO at Phala.

A Broader Shift in Enterprise AI

The trend extends far beyond law. Healthcare providers, banks, and insurers are adopting confidential compute to meet data-protection standards while using generative AI for research, analytics, and claims processing. Analysts expect the global confidential computing market to exceed $70 billion by 2030, driven by increased regulatory scrutiny and demand for verifiable privacy.

Recent guidance from the European Data Protection Board (EDPB) also supports this direction, calling for “computational locality” and traceable data governance in AI systems. Confidential computing delivers those guarantees by ensuring that sensitive workloads remain sealed, verified, and geographically contained.

About Phala

Phala is a privacy infrastructure company building the foundation for confidential AI and data processing. By leveraging hardware-secured enclaves and blockchain-based attestation, Phala enables organizations to process sensitive workloads in a verifiable, privacy-preserving way. The company’s enterprise platform integrates with standard cloud and software ecosystems to bring confidential computing to regulated industries worldwide.

Learn more at phala.com.

Fonte: Business Wire