AI Tool Risk and Cost
Most lawyers using ChatGPT, Claude, Gemini, Perplexity, or Grok at work are on consumer-tier accounts with no admin visibility, no audit log, and inputs that may train the next model. Moving to an enterprise tier closes the confidentiality gap. It does not close the hallucination gap, which is what gets attorneys sanctioned.
This page shows the risk profile and the price of each tier, so a firm administrator can size the trade-off: stay on consumer plans and rely on a written verification policy, or pay several thousand dollars per lawyer per year for the audit log an enterprise tier provides.
On this page
At a glance
- 790 court, disciplinary and administrative matters addressing AI use tracked, 390 of them decided matters the tracker codes as a court sanction or bar discipline.
- The tracker records the AI tool as free text, often with the court's own qualifications, so this page does not count matters by tool. Each card below links to the tracker entries that mention the product.
- Median stated monetary sanction: $2,000 across 179 sanctioned matters with a dollar amount.
Risk and cost by tier
Up to three tiers per vendor: consumer plans (personal accounts), a multi-user mid-tier where one exists, and the contracted enterprise tier. Annual cost is per user, list price, before negotiation. Vendor pricing verified 2026-05-04 (ChatGPT, Claude, Gemini, Microsoft Copilot) and 2026-05-06 (Perplexity, Grok).
| Plan tier | Risk | Annual cost / user | Notes |
|---|---|---|---|
| ChatGPT Tracker entries that mention ChatGPT → | |||
| Consumer (Free, Go, Plus, Pro) | High | $0 to $1,200+ | Personal accounts. OpenAI may train on inputs unless opted out. No admin visibility, no audit log. |
| Business (multi-user) | Medium | $240 to $300 (2-seat minimum) | No training on inputs. SOC 2 Type 2. No Compliance API access on this tier. |
| Enterprise | Low | Sales-contracted. OpenAI does not publish a price floor or seat minimum. | SOC 2. Compliance Logs Platform unified release March 2026, replacing the legacy Compliance API. |
| Claude Tracker entries that mention Claude → | |||
| Consumer (Free, Pro, Max) | High | $0 to $1,200+ | Personal accounts. Anthropic consumer terms; review current data-handling settings. |
| Team (Standard, Premium) | Medium | $240 to $1,500 (5 to 150 seats) | Multi-user workspace. No Compliance API on Team tier. |
| Enterprise | Low | $240+ per seat plus API usage (annual contract) | Compliance API GA 30 March 2026 (Enterprise plan, excludes Public Sector). Captures admin and resource activity. |
| Gemini (Google) Tracker entries that mention Gemini → | |||
| Consumer (Google AI Pro, Google AI Ultra) | High | $240 to $3,000 | Personal accounts under Google consumer terms. Renamed from "Gemini Advanced." |
| Workspace Business (Standard, Plus) | Low | $168 to $264 (Gemini bundled into base SKU) | Admin SDK Reports API exposes Gemini activity. 180-day rolling log history (records from June 2025 forward). |
| Microsoft Copilot Tracker entries that mention Copilot → | |||
| Microsoft 365 Premium (consumer) | High | $240 | Replaces the deprecated Copilot Pro standalone. Personal account, consumer terms. |
| Microsoft 365 Copilot (enterprise add-on) | Low | $360 add-on plus base license: $150 (Business Standard) to $684 (E5). Total: $510 to $1,044. | Requires Microsoft 365 Business Standard, Business Premium, E3, or E5. Purview audit logging included; advanced retention requires E5 or the Purview Audit (Premium) add-on. Microsoft Graph aiInteractionHistory API captures prompts, accessed resources, and responses. |
| Perplexity Tracker entries that mention Perplexity → | |||
| Consumer (Free, Pro, Pro Max) | High | $0 to $2,004 | Personal accounts. Pro at $204/yr; Pro Max at $2,004/yr is the highest-priced consumer LLM tier on this page. Forest Ridge v. Heag (N.C. Ct. App. 2026) involved Perplexity.AI Professional and a public chastisement of counsel. |
| Enterprise Pro | Low | Sales-contracted. Perplexity does not publish a per-seat price or seat minimum. | SOC 2 Type 2. Enterprise customer data excluded from model training. SSO and SCIM. Configurable file retention. A dedicated audit-log API is not documented on the enterprise marketing page. |
| Grok (xAI) Tracker entries that mention Grok → | |||
| Consumer (Free, SuperGrok Lite, SuperGrok) | High | $0 to $360 | Personal grok.com accounts. xAI consumer terms; "Private Chat" is the consumer opt-out. X Premium and Premium+ ($32 to $395/yr) bundle Grok access on a separate consumer surface. Billups v. Louisville Mun. Sch. Dist. (N.D. Miss. 2025) sanctioned an attorney for using Grok despite a firm policy against external AI. |
| Grok Business | Medium | $360 | Multi-user team workspace. Excluded from model training by default. Centralized billing, user analytics. No published audit-log API at this tier. |
| Grok Enterprise | Low | Sales-contracted. xAI does not publish a per-seat price or seat minimum. | SOC 2 Type 2 per xAI Enterprise. SSO, SCIM, custom role-based access controls, custom data retention. Audit logging documented on the xAI Enterprise page. |
Cost of risk vs. cost of mitigation
Set the price of an enterprise tier alongside the price of a single sanction event and the trade-off becomes legible. Sanctions in our tracker run from $500 to seven figures. Bar discipline ranges from public admonishment to suspension. Only one of the malpractice application forms reviewed for this site asks about AI, but broker commentary reports LPL underwriters focusing on firms' AI controls.
If a sanction lands
- $500 to $10,000+ direct monetary sanction (modal range in tracker)
- Conditions on future appearances, such as the two years of mandatory co-counsel in the district ordered with a $1,000 fine (Hatfield v. Pirani, W.D. Ark.)
- $110,204 in sanctions and fee award (Couvrette v. Wisnovsky)
- Plus pro-hac vice revocation, bar referral, fee disgorgement
- Plus malpractice carrier renewal questionnaire flag
- Plus public docket entry, news coverage, client confidence loss
If a written policy lands first
- Carrier-Renewal Packet: tiered pricing, one-time fee. Policy, training, and vendor diligence captured before the next renewal questionnaire.
- Consulting: 4 to 8 week productized engagement. Vendor diligence, live training rollout, written policy with the firm's tool tiering captured.
- Less than the cost of one Rule 11 sanction. Less than the annual cost of upgrading two attorneys to ChatGPT Business.
Sources
- OpenAI, ChatGPT pricing (verified 2026-05-04): openai.com
- OpenAI, ChatGPT Enterprise: openai.com
- OpenAI, Compliance APIs (March 2026 unified Compliance Logs Platform): help.openai.com
- OpenAI Trust Center (SOC 2 Type 2): trust.openai.com
- Anthropic, Claude pricing (verified 2026-05-04): claude.com
- Anthropic, Compliance API (Claude Enterprise GA 30 March 2026): support.claude.com
- Google, Workspace pricing (verified 2026-05-04): workspace.google.com
- Google, Gemini consumer subscriptions (verified 2026-05-04): gemini.google
- Google, Gemini audit logs in Reporting API (July 2025 release): workspaceupdates.googleblog.com
- Google, Admin SDK Reports activity reference, gemini_in_workspace_apps: developers.google.com
- Microsoft 365 Copilot for business (verified 2026-05-04): microsoft.com
- Microsoft 365 enterprise pricing (verified 2026-05-04): microsoft.com
- Microsoft, Audit Copilot interactions (Microsoft Purview): learn.microsoft.com
- Microsoft, aiInteractionHistory resource (Microsoft Graph): learn.microsoft.com
- Perplexity, Pro pricing (verified 2026-05-06): perplexity.ai
- Perplexity, Enterprise: perplexity.ai
- Perplexity Trust Center (SOC 2 Type 2): trust.perplexity.ai
- xAI, Grok plans (verified 2026-05-06): grok.com
- xAI, Enterprise (SOC 2 Type 2, audit logging): x.ai
- xAI, Grok Business documentation: docs.x.ai
- ABA 2024 Legal Technology Survey Report, AI TechReport: americanbar.org
Vendor pricing and risk profile verified 2026-05-04 (ChatGPT, Claude, Gemini, Microsoft Copilot) and 2026-05-06 (Perplexity, Grok). Tracker counts updated at every build. AI tool pricing changes frequently; reverify before policy decisions.