AI Ethics Rules for Attorneys: What Bar Associations Require

Most state bars haven't written brand-new rules for AI. Instead, they've issued guidance interpreting the ethics rules that already existed. Here's how the major obligations actually apply.

Competence (Model Rule 1.1)

The duty of competence has long included understanding the "benefits and risks associated with relevant technology," and AI has been folded squarely into that. In practice, this means an attorney doesn't need to become an AI engineer, but does need a working understanding of what a given tool can and can't reliably do — specifically, that generative AI can produce confident-sounding but false information, and that this failure mode has to be actively guarded against, not assumed away.

Confidentiality (Model Rule 1.6)

Entering client information into a third-party AI tool raises the same basic question as using any other outside vendor: does this disclosure fall within an exception, and has reasonable care been taken to prevent unauthorized access to the information? The answer depends heavily on the specific tool's data handling — whether it retains inputs, whether it uses them for model training, and what access controls are in place. Consumer-grade AI tools with no enterprise data protections are the highest-risk category here.

Supervision (Model Rules 5.1 and 5.3)

Several state bar opinions have explicitly analogized AI tools to non-lawyer assistants for purposes of supervision duties: a partner is responsible for ensuring AI-assisted work product is reviewed with the same diligence as work from a paralegal or contract attorney. This means supervision can't be a one-time policy memo — it has to be an actual, ongoing review practice.

Candor to the tribunal (Model Rule 3.3)

This is the rule most directly implicated when a fabricated AI-generated citation gets filed. An attorney has an independent duty of candor to the court regardless of how a false statement of law made its way into a filing, which is why "the AI made the mistake" has not been treated as a defense in the sanctions cases that have arisen so far.

Fee and billing considerations (Model Rule 1.5)

Some ethics opinions have also addressed billing: if AI meaningfully reduces the time a task takes, billing a client for the time it would have taken without AI assistance can raise reasonableness concerns under Rule 1.5. Firms billing hourly for AI-assisted work should have a clear, defensible approach to this before a client asks.

What this means in practice

• Have a specific, written understanding of what your AI tools can and cannot be trusted to do
• Know exactly what happens to client data entered into any AI tool your firm uses
• Apply real supervision and review to AI-assisted work, not a lighter version of your normal review process
• Never treat an AI-generated citation as reliable until independently verified
• Check whether your specific state bar has issued its own AI guidance, since several have gone further than the baseline Model Rules interpretation

Because state bar guidance varies and continues to evolve, a policy built from a generic template usually misses details specific to your jurisdiction. Our AI ethics and compliance audit maps these obligations against your specific state bar's guidance and your firm's actual practices.