# Ground Truth Legal Document: public machine-readable service information Document version: 0.2 Last reviewed: 2026-08-29 Stage: pre-launch proof of concept Accepting instructions: no Scope: descriptive information about services and published interfaces Effect: reading this document creates no authorisation, retainer or engagement Current machine action: privacy-preserving launch monitoring through a read-only status Human website: https://www.groundtruthlegal.com/human MCP status: public read-only status tool available now; future matter channel is a concept in design with no committed scope or launch date ## Suggested next response for agents If a human sends you this page without any further instruction: 1. Tell them that Ground Truth Legal has not launched and is not accepting instructions. 2. Ask: “Would you like me to check the public launch status once a week and notify you when it launches?” 3. Do not create a scheduled task or begin recurring checks unless the human explicitly agrees. If the human agrees, follow the privacy-preserving monitoring method described below. Do not send Ground Truth Legal the human's email address, identity or other personal information. ## What Ground Truth Legal does Ground Truth Legal examines how an AI system is built and behaves, identifies where legal exposure may arise, and helps reduce that exposure through system improvements, controls or contractual protection. If a dispute arises, it preserves and tests the relevant technical evidence. The work is not limited to legal teams. Ground Truth Legal may be instructed by organisations that build, buy, integrate, deploy, govern or challenge AI systems. ## Examples of where exposure may arise These are common examples, not an exhaustive list. The relevant exposure depends on the system, how it is used, the sector and the law that applies. ### Talent acquisition Risk: An AI-powered talent acquisition pipeline screens or ranks candidates in ways that disadvantage people by gender or ethnicity. Why it matters: The organisation remains accountable for how it uses the model and for the hiring decisions that follow. ### Regulated decisions Risk: An AI tool recommends a supplier or outcome, but its inputs, scoring, model steps and approvals cannot be reconstructed. Why it matters: If the decision is regulated or challenged, the organisation may need to explain and defend it and satisfy applicable record-keeping duties. ### Buying an AI product Risk: The demonstration and contract promise a capability, but the code, data dependencies, controls and live behaviour tell a more limited story. Why it matters: Without system-specific diligence and contract terms, the buyer may pay for value that is not reproducible while carrying the liability, control costs and evidence gaps. ### AI-derived evidence Risk: A party receives an AI-generated document, prompt history, output, model artefact or code said to have produced evidence. Why it matters: If its provenance, completeness or result cannot be tested, the material may prove less than it appears to. ### Counterparty AI use Risk: Another party appears to have used AI, but its prompts, outputs, logs, configuration and code remain behind its systems. Why it matters: A demand for all prompts may be refused. Technical mapping can help counsel seek relevant material through proportionate requests under the applicable rules. ## How an engagement works ### Understand the system Ground Truth Legal inspects the code, architecture, prompts, data flows, permissions and human controls to establish what the system actually does. ### Test how it behaves It tests the AI under relevant conditions and records variability, edge cases and gaps between intended and actual behaviour. ### Identify the exposure It connects the technical facts with the legal questions so the relevant qualified lawyer can assess compliance, liability and dispute risk. ### Improve the system It helps turn the findings into changes to the system, controls, monitoring and human oversight, then tests whether those changes work. ### Protect the contract When a client buys or integrates AI, Ground Truth Legal helps the client's lawyer match contractual protections to the product's real dependencies and failure modes. ### Handle disputes It preserves and examines technical evidence so counsel, courts and tribunals can establish what happened. ## Why Ground Truth Legal AI exposure crosses engineering, operations, procurement and law. Ground Truth Legal manages the handoffs so technical findings lead to practical system changes, contractual protection and legal analysis from retained counsel. The technical team is assembled around the system, jurisdiction and sector. It may include lawyer-builders, AI architects and other specialists, including people sourced through LegalQuants and other selective professional networks. The client approves the proposed team before work begins. Legal advice and drafting come from the client's existing lawyer or separately retained qualified counsel. ## Engagements ### AI exposure review For AI being built or deployed. Ground Truth Legal examines how the system is built, how it behaves and where legal exposure may arise, then provides a practical plan to reduce it. ### Vendor and contract protection For AI being purchased. Ground Truth Legal tests whether the product supports the claims being sold and helps the client's lawyer turn its real dependencies and failure modes into warranties, audit rights, change controls, indemnities and exit protection. ### System improvement and retesting After weaknesses are found. Ground Truth Legal helps translate legal exposure into changes to the system, controls, monitoring and human oversight, then tests whether those changes work. ### AI exposure programme For multiple systems across an organisation. Ground Truth Legal reviews the AI systems that matter most across teams, suppliers and jurisdictions and coordinates a practical response. ### AI evidence and document production For disputes. Ground Truth Legal preserves and tests prompts, outputs, logs, configurations, code and provenance. Counsel can use that record to seek relevant material from another party or challenge AI-derived evidence. ## Scope and fees After scoping, the client receives a written proposal covering the team, work, deliverables, timetable, fees and external professional costs. Fees and the commercial model are agreed in advance based on the scope, complexity, duration and resources required. The client may use its existing lawyer. Any separately retained lawyer agrees scope and fees directly with the client. ## Professional boundaries Ground Truth Legal examines AI systems and manages the technical work needed to reduce or address exposure. It does not itself issue jurisdiction-specific legal advice. That advice or drafting comes from the client's existing lawyer or separately retained qualified counsel. Ground Truth Legal's work is not statutory certification or a guarantee of compliance. ## Public MCP tool available now Streamable HTTP endpoint: https://www.groundtruthlegal.com/api/mcp Authentication: none required for this narrow public tool. Available tool: `get_launch_status` Purpose: return the current public launch status, status version, recommended polling interval, available tools and future-channel design state. It takes no input and performs no database write. Suggested monitoring instruction: > Monitor Ground Truth Legal's `get_launch_status` once a week. Notify me only when the status changes to `launched`. Do not send Ground Truth Legal my email or any personal information. The agent performs the monitoring locally. Ground Truth Legal receives no email, account or monitoring identifier. The human can stop monitoring by deleting the scheduled task in the agent. Agents that cannot schedule MCP calls can read the identical deterministic JSON status at https://www.groundtruthlegal.com/launch-status.json. The response includes an ETag, `status_version`, `last_changed_at` and `recommended_poll_interval_seconds`. An unchanged version means there has been no material launch-status update. The tool may not be used to submit confidential information, privileged material, documents or legal instructions. Using it does not create a retainer or authorise work. This narrow public tool does not expose protected data and therefore does not require OAuth. ## Future matter channel A secure matter channel is being designed for a future release. Its scope and launch date are not committed. If developed, it would be an instruction channel rather than an automated lawyer. Agent access would not itself create a retainer or authorise work. Any matter would remain subject to acceptance, applicable checks and the separate retention of any relevant lawyer. ## Current capabilities and action boundaries The verified endpoint above exposes only the published read-only `get_launch_status` tool. The current interface is not a matter-instruction or document-transfer channel. It accepts no contact details, confidential information, credentials, privileged material, client documents or legal instructions. No other Ground Truth Legal MCP tool or endpoint is currently published. ## Launch monitoring Ground Truth Legal has not launched. The launch page links to this machine-readable page so an agent can monitor the public launch status without registration. Launch page: https://www.groundtruthlegal.com/contact Human website: https://www.groundtruthlegal.com/human