AI ETHICS and customer support. The policy should specify who may commit the office to AI vendor relationships and what approval process applies, which requires a careful review of any third-party service terms either directly imposed by or flowed down from underlying model providers. This is particularly important because upstream obligations are often unenforceable by the attorney yet fully enforceable against the firm or office, creating an asymmetric risk exposure that most users do not recognize. Solo practitioners often make these decisions personally; larger organizations need procurement procedures that ensure appropriate review before adoption. Offices committed to sustainability should prefer vendors with demonstrated environmental practices, including renewable energy commitments and carbon and/or water reduction goals.
8.
Disclosure and Transparency
Many courts now require disclosure or certification regarding AI use in filings. NACDL acknowledges this and strongly believes that court rules about disclosure and transparency should apply equally to both the defender and prosecution functions (and their various agents). With that caveat, policies should require compliance with all applicable court orders regarding AI disclosure, establish procedures for tracking and certifying AI use, identify responsibility for monitoring new disclosure requirements, and create documentation practices that support accurate certification. For example, the policy might have specific language the firm or office uses for AI disclosures, either as one single disclosure or as multiple models for different uses or outputs. Some offices may choose to adopt disclosure practices beyond what courts require. Voluntary transparency about AI use can demonstrate good faith and preempt concerns that might otherwise arise.
9.
Monitoring and Enforcement
Policies without enforcement become suggestions. Effective policies establish mechanisms for monitoring compliance, auditing AI use against policy requirements, investigating potential violations, and imposing appropriate consequences for violations. Monitoring does not have to be intrusive. Technical controls can restrict access to approved tools and flag potentially problematic uses, although this often requires the vendor to configure the monitoring unless the firm or office has sophisticated information technology support or other qualified professionals. Review of work product can identify AI-related issues. Perhaps most important, a culture of compliance in which defenders and staff view ethical and efficient AI use as a shared professional responsibility reduces the need for heavy-handed enforcement. When violations occur, proportionate responses maintain policy credibility. Minor inadvertent violations might warrant coaching and remediation. Serious or repeated violations might justify formal discipline. The policy should provide enough guidance that personnel understand the stakes without creating rigidity that discourages using AI altogether.
44
Parity in Practice: The Defender’s Duty to Ethically Use AI
Made with FlippingBook Online document maker