Anthropic Legal Bearings Pack AI governance & counsel workflows
Overview · Navigation

Counsel Work

Overview

Counsel workboard for Anthropic integrations

S5 turns the narrative from S1–S4 into a concrete, repeatable workflow for live matters. It is a checklist for counsel: what to map, what to check, what to design, and when to revisit a scenario.

Phase 1 · Map the use-case and actors

  • Define the system boundary and the specific scenarios you care about.
  • Identify who holds which levers (Anthropic, your client, partners, end users, regulators) using S2.
  • Flag whether the scenario is primarily safety/misuse, privacy/autonomy, or “ordinary” productivity and workflow.

Phase 2 · Align with Anthropic’s posture

  • Use S1–S3 and the policy handouts to understand Anthropic’s safeguards, governance, and usage boundaries for the relevant products.
  • Pull specific Anthropic policies, RSP / ASL materials, and evaluation sections into your evidence shelf.
  • Note any gaps between what your scenario needs and what Anthropic currently commits to.

Phase 3 · Design controls, terms, and UX

  • Design technical, organisational, and contractual controls that bridge the gap between Anthropic’s posture and your client’s risk appetite.
  • Use S4 and the Tools & Templates to shape disclaimers, allocation of responsibility, and governance processes.
  • Draft or update internal policies so that what the system does, what the contract says, and what teams are trained to do all line up.

Phase 4 · Land lanes and plan re-checks

  • For each scenario, record a lane decision in S6 (Green / Amber / Red), with evidence pins and follow-ups.
  • Set explicit review triggers: time-based (e.g. 6–12 months) and event-based (major incidents, model updates, policy changes).
Counsel ownership

This workboard is a scaffold. It does not decide what is lawful or acceptable. Human counsel owns each phase, each lane, and each recommendation, even when using templates and checklists.

Template

Matter-level checklist

Use this checklist when you open a new matter. You can copy this table into your own systems and expand it, but keep the “Evidence noted?” and “S6 recorded?” checkpoints.

Phase Key tasks Primary bundles / sources Owner(s) Evidence noted? S6 recorded?
1 · Map Define scenarios and system boundaries;
map actors and levers; flag safety / privacy hotspots. 		S1, S2, Foreseeable Misuse index, Penumbral spine. e.g. Product counsel + product lead Y / N + where recorded N (too early) / partial / initial
2 · Align Pull Anthropic policies, RSP / ASL materials, and model evaluations; identify doctrinal tracks and key uncertainties. S3, S4, Policies & handouts, Reading Stack. e.g. Product counsel + regulatory specialist List source IDs / URLs Optionally draft provisional lane(s)
3 · Design Design controls, disclaimers, and governance; align internal policies and contracts with how the system will behave. Tools & templates, client playbooks, internal policies. e.g. Product counsel + commercial / privacy Attach draft terms / policies Lane ready for sign-off?
4 · Land & revisit Record final lanes in S6; set review triggers; ensure incidents and changes feed back into governance. S6, governance records, incident / audit logs. e.g. Product counsel + risk committee Confirm evidence archive location Final lane date + approver
Interpretive transparency

This checklist assumes you will record which Anthropic materials, legal sources, and internal artefacts you relied on. That record is part of your defence posture if a lane is later questioned.