T

Truth Systems

The compliance agent Truth Systems describes reads a firm's policies and its clients' guidelines and turns each mandate into live access controls, matter by matter, and the dashboard classifies AI use by use case. Without the agent, a browser extension, desktop agent, rule engine, dashboard and audit log remain, which is most of what a firm deploys. Every block or warning a lawyer sees depends on how the agent read a client's guidelines. Alexandria applies the same checks to AI agents, with every tool call checked against the agent's identity.

Truth Systems says Charter catches what crosses the line before the damage is done, and a product mockup on its home page counts 276+ prevented incidents, with no method or named firm attached. Nothing published says how often the agent misreads a client guideline, how often a prompt that should have been stopped gets through, or how often a legitimate prompt is blocked. The product cites no legal authority, so grounding takes a different form here: the rules it enforces come from its own reading of client guidelines, and nothing says whether each rule links back to the passage it was drawn from.

Charter checks each prompt before it leaves the browser and blocks it or warns the user, and Truth Systems says it turns each client mandate into live access controls, matter by matter. A dashboard and audit log let the firm see what was blocked. No review step over the agent's work is published: nothing says whether a person confirms the rules the agent derives from a client's guidelines before they are enforced, what happens when a guideline is ambiguous, or how a lawyer can challenge a block on a live matter.

Alexandria's agents are described as acting under a real identity, with every tool call checked against it, and no human review step is described.

The home page carries quotes from three people at Arnall Golden Gregory, alongside a customer story video. Sean Fogarty, a managing partner, says what Charter gives the firm is transparency. Jennifer Waite, chief knowledge and innovation officer, says clients want to know exactly how the firm works with AI on their matters: which tool was used, what it did and how much time it saved. Alexander Hughes, director of information governance, says the firm is not worried about data leakage because it monitors, watches, stops and redirects it.

At its August 2025 launch the vendor said it worked with two law firms with $2 billion in combined revenue and nearly 1,500 attorneys. Product mockups on the home page show counts such as 276+ prevented incidents and a 51 percent rise in hours spent on AI, with no named firm or method. No figure with a method is published: how many incidents were prevented at a named firm, over what period, and against what baseline.

Charter captures real sessions, so the vendor's system holds privileged material from every AI interaction it monitors. Access controls, custom retention, single tenant architecture and deployment on premises are offered. Nothing says whether captured prompts train any model, whether a model provider processes them when a prompt is evaluated, or how blocked privileged content is itself stored. No terms of service, privacy policy or security page is published.

Nothing published says where the product's judgment stops and the firm's begins. There are no terms of service or usage policy on the site. The product gives clients no legal advice, but it does interpret a law firm's obligations: its agent reads each client's guidelines and decides which prompts to allow, and nothing says the firm remains responsible for that reading or should confirm it. The product is sold to law firms.

No governance position is published for the vendor's own AI. The product governs other vendors' AI tools, and nothing describes how Truth Systems governs its own: who is accountable for the agent's reading of client guidelines, what is tested before a release, or how errors and uneven results are found and disclosed. A firm relying on Charter to enforce its clients' mandates relies on that agent to read them correctly.

Truth Systems' home page names single sign on, IP allow listing, granular access by role, and data lifecycle and retention controls. No privacy policy, subprocessor list, deletion commitment, encryption statement or incident practice is published, for a system that captures prompts and responses from every AI tool a firm uses.

Nothing is published on who bears the loss when the product is wrong. No terms of service or customer agreement is published, so no warranty, indemnity or liability cap can be read before signing. If Charter lets a privileged prompt through, or misreads a client's ban on AI, the breach lands on the firm and its client.

Firm systems are named and none of the connections is documented. Alexandria is described as an MCP gateway connecting Intapp, iManage, Aderant and Microsoft 365 into one layer for AI agents, used for conflicts clearing, intake and matter setup, with every agent acting under a real identity and each tool call checked against it. Charter runs as a browser extension and desktop agent across whatever AI tools a firm uses, with Claude and Harvey shown in the product imagery, and client guidelines are uploaded as PDF or DOCX files.

Nothing describes what each connection reads, in which direction or what a firm must configure. No integrations page or developer documentation for either product is published; the vendor's documentation site covers only an earlier hallucination detection SDK called Gateway.

Truth Systems' home page lists deployment on premises, single tenant architecture and geolocation control, so the tenancy model is stated. It names no hosting provider or region, does not say which regions geolocation control covers, and does not say where prompts are processed when Charter evaluates them, as distinct from where captured sessions are stored.

No independent security attestation is published. The home page describes a robust security program and lists access features, but names no SOC 2 report, ISO certification, penetration test or trust center, for a product that captures every AI prompt and response in a law firm.

A compliance AI agent is referred to without saying what runs underneath it. In October 2025 cofounder Nam Nguyen said the product uses a blend of smaller language models and large language models, and that the company wants smaller, more predictable models rewarded for spotting contradictions. No model, model provider or hosting arrangement for inference is named, and nothing commits to notice when that changes. Whether a lawyer's prompt is evaluated by a model, and whose, is not stated, so a firm cannot tell its clients where their most sensitive text goes when Charter checks it.

No pricing is published at any level: no price, no tier and no unit of charge. The route to a price is a demo request.

The site addresses law firms throughout, the named customer is a law firm, and cofounder Nam Nguyen has said the company mainly targets the AmLaw 200; its Y Combinator profile also names banks, financial services and other regulated industries, and lists a team of six. Charter covers human use of AI across tools and Alexandria covers AI agents, with reporting by client and matter. Not stated: which browsers, desktop systems and AI tools Charter supports, whether Alexandria is generally available, and what firm size the product is built for.