Skip to content
Approved tools

What’s cleared, what isn’t.

For each tool: what it’s good at, what it isn’t, the pros and cons, and the committee’s reasoning for approval. New term? Check the glossary.

Claude

Approved · Tier 1-4
by Anthropic
Approved for
Tier 1-4 (Claude Enterprise seat holders)

The company's AI tool: the Enterprise plan, with our data protections and limited seats

Visit Anthropic
Best for
  • Internal and confidential work (Tier 2-4) for company seat holders
  • Long-document work: upload contracts, policies, or reports and work through them in place
  • Recurring workflows: Projects keep files and instructions together so repeat work does not start from scratch
  • Drafts where the deliverable matters, so outputs become editable, shareable documents instead of chat text
Not for
  • Customer or applicant data on a free or personal login. It only belongs on your company Enterprise seat
  • Work without a seat: free claude.ai is a consumer product, Tier 1-2 only under the free-tool lane
  • Final review. A human verifies anything that leaves the building
Pros
  • Company-managed Enterprise plan: contractual no-training on our content, signed DPA, SSO, and audit logs
  • Independently audited security and AI-management certifications on the commercial plans
  • Built for document work: upload PDF, Word, and Excel files and interrogate them directly
  • One company agreement covers web, desktop, and mobile use under the same protections
Cons
  • Free claude.ai is NOT covered by the company plan. Consumer terms apply, so it stays in the free-tool lane (Tier 1-2, with guardrails)
  • Like every AI tool, it can state wrong facts with confidence, so verify outputs before relying on them
  • Conversations are processed in the vendor's cloud, and retention is governed by plan settings, not by deleting later
Why the committee approved this

The committee's recommended standalone AI tool. The company's Claude Enterprise plan contractually excludes training on our content, carries a signed DPA, SSO, and audit logging, and the commercial plans are independently audited. That is what clears it for Tier 1-4 work for seat holders, including customer data (Microsoft 365 Copilot seats carry the same scope for those who have them). No seat? Free claude.ai falls under the free-tool lane (Tier 1-2, with guardrails). Customer and applicant information only ever goes on a company enterprise seat, never a free or personal tool. Facts re-verified against the vendor's published terms at each 60-90 day review.

Official links
VendorEnterprise planTrust centerPricing

Microsoft Copilot

Approved · Tier 1-4
by Microsoft
Approved for
Tier 1-4 (Microsoft 365 Copilot seat holders)

Microsoft's AI assistant: company-protected inside Microsoft 365 for seat holders, consumer terms everywhere else

Visit Microsoft
Best for
  • Working inside the Office apps with a company seat: drafting in Word, summarizing in Outlook and Teams
  • Internal and confidential work (Tier 2-4) for Microsoft 365 Copilot seat holders
  • Quick public-info tasks in the consumer version when you don't have a seat
Not for
  • Customer or applicant data on the CONSUMER Copilot. It only belongs on the company-seat version
  • Assuming 'Copilot' is one product. The name covers several with different data terms
  • Final review. A human checks anything customer-facing
Pros
  • Company Microsoft 365 Copilot seats keep work data inside the company tenant under Microsoft's business protections
  • Lives inside the Office apps people already use, so there is no separate tool to learn
  • The consumer version is free for public and general internal work under the free-tool lane
Cons
  • Two very different products share the name: a company seat is protected, the consumer app is not, so check which one you're in before pasting anything
  • Consumer Copilot runs on personal-account consumer terms: no DPA, no admin controls, no audit trail
  • Like every AI tool, it can state wrong facts with confidence, so verify outputs before relying on them
Why the committee approved this

Some employees hold Microsoft 365 Copilot seats through the company's Microsoft licensing. Those seats keep work data inside the company tenant under Microsoft's business data protections, so seat holders are approved for Tier 1-4 work, the same scope as Claude Enterprise seats. Everyone else: the consumer Copilot falls under the free-tool lane, Tier 1-2 (public and general internal work), with guardrails. The protections come from the company seat, not the brand name, so check which Copilot you're in before pasting anything.

Official links
VendorMicrosoft 365 Copilot (business)Trust center

ChatGPT

Approved · Tier 1-2
by OpenAI
Approved for
Tier 1-2 (free-tool lane)

A widely known AI assistant, on consumer terms unless a company buys the business plan

Visit OpenAI
Best for
  • Public-info drafting and brainstorming when it's the tool you already know
  • Learning AI skills and practicing prompts, since familiarity makes it the easiest on-ramp
  • Quick rewording and formatting of non-sensitive text
Not for
  • Customer or applicant information: never in a free or personal tool; that needs a company enterprise seat
  • Company financials, pricing, underwriting, or security details: those are Tier 3, company seat only
  • Assuming a personal paid subscription adds protection: it doesn't, and consumer terms still apply
Pros
  • Free tier available to everyone, with no seat or setup needed
  • The easiest AI-literacy on-ramp because so many people already know it
  • OpenAI's business plans do exclude training by default, relevant only if the company ever buys them
Cons
  • Free and personal-paid tiers may use conversations for training unless you opt out in settings, and opting out only covers future chats
  • Feedback buttons (thumbs up/down) can put a conversation back into training even after opting out
  • Familiarity breeds overconfidence: 'everyone uses ChatGPT' is not a data protection
Why the committee approved this

Advantage+ has not purchased a ChatGPT business plan, so every ChatGPT account an employee can access runs on consumer terms. That places it in the free-tool lane: Tier 1-2 (public and general internal work), with guardrails. Turn off training in settings, avoid the feedback buttons on anything work-related, and verify outputs. A personal paid subscription does not change the data relationship.

Official links
VendorBusiness plansHow your data is usedSecurity

Gemini

Approved · Tier 1-2
by Google
Approved for
Tier 1-2 (free-tool lane)

Google's AI assistant that fits naturally inside the Google ecosystem, with consumer terms on personal accounts

Visit Google
Best for
  • Public-info research and drafting when your personal workflow is Google-centric
  • Mobile-first quick questions, built into Android with solid apps
  • Quick web-grounded answers when you don't need checkable citations
Not for
  • Customer or applicant information: never in a free or personal tool; that needs a company enterprise seat
  • Company financials, pricing, underwriting, or security details: Tier 3, company seat only
  • Anything you'd mind a human reviewer reading, since Google states a sample of consumer chats is human-reviewed
Pros
  • Free and broadly available, including on phones
  • Can ground answers in Google Search for current public information
  • Google's Workspace version carries business data protections, relevant only if the company ever adopts Google Workspace
Cons
  • Consumer chats are used for training by default, and a sampled subset is read by human reviewers, whose reviewed chats are kept for years even if you delete your activity
  • Turning activity-saving off still leaves a short retention window, and feedback can re-open the training path
  • The Gemini name covers many products with different data terms, and the brand alone never tells you which protections apply
Why the committee approved this

Advantage+ does not run on Google Workspace, so there is no company-managed Gemini. Consumer Gemini falls under the free-tool lane: Tier 1-2 (public and general internal work), with guardrails. Turn off activity-saving first, and follow Google's own guidance not to enter anything confidential. The Workspace business tier would warrant committee review only if the company ever adopted Google Workspace.

Official links
VendorGemini for Workspace (business)Consumer privacy hubWorkspace AI privacy

Perplexity

Approved · Tier 1-2
by Perplexity AI
Approved for
Tier 1-2 (free-tool lane)

Search-first AI: answers built from live web results, with citations you can check

Visit Perplexity AI
Best for
  • Research on current public topics: rates, regulations, vendor news, market moves
  • Questions where you need to verify the source, not just trust the answer
  • Getting oriented on an unfamiliar company or industry quickly
Not for
  • Customer or applicant information: never in a free or personal tool; that needs a company enterprise seat
  • Company financials, pricing, underwriting, or security details: Tier 3, company seat only
  • Polished long-form drafting: it is built to find and cite, not to write deliverables
Pros
  • Inline citations let you click through and check where every claim came from
  • Grounded in the live web, so it suits fast-moving questions where static knowledge lags
  • The free tier covers Tier 1 research work without any purchase
Cons
  • Training-style data collection is ON by default for free AND personal-paid accounts. Opt out in settings while logged in, and it only covers future queries
  • Paying for the personal Pro plan does not add company protections. The data relationship changes only on an enterprise contract, which the company has not purchased
  • A citation proves where a claim came from, not that the source is right, so verify the links that matter
  • Answers change as the web changes, so do not rely on re-running a search to justify a past decision
Why the committee approved this

Advantage+ has not purchased Perplexity's enterprise tier, so all employee access runs on consumer terms. That places it in the free-tool lane, approved for Tier 1-2, though its real strength is public research. Log in and turn off the data-retention toggle in settings before any work use. Its citation-first design makes it a strong free option for verifiable public research, which is exactly the lane it is approved for.

Official links
VendorEnterprisePrivacy policyTrust center

Free AI Tools

Approved · Tier 1-2
by Various
Approved for
Tier 1-2 (public + general internal, with guardrails)

Every consumer AI tool, as a category. Approved for public and general internal work, with guardrails

Best for
  • Drafting public-facing copy from public information
  • Brainstorming, idea generation, and learning AI skills
  • Summarizing public articles and industry research
  • Anything you would be comfortable seeing on a public website
Not for
  • Customer or applicant information: never; that only belongs on a company enterprise seat
  • Company financials, pricing, underwriting, or security procedures: Tier 3, company seat only
  • Anything you wouldn't want a stranger, or a human reviewer at the vendor, to read
Pros
  • Free and available to everyone, with no seat required
  • Keeps the whole company building AI literacy, not just seat holders
  • Good enough for public-data drafting, brainstorming, and research
Cons
  • Most consumer tools use your chats for training by default. Opt-outs exist but live in per-account settings, only cover future chats, and can be undone by feedback buttons
  • Some vendors have humans review a sample of conversations, and reviewed chats can be retained for years, even after you delete them
  • No DPA, no admin controls, no audit trail, so the company has no visibility into a personal account
  • Settings are per-person and per-account, so the company cannot verify a colleague's toggles
Why the committee approved this

Most of the company won't have an enterprise seat, and banning AI outright would push usage into the shadows. So the committee's stance is a two-lane policy: free tools are approved for Tier 1 and Tier 2 (public and general internal work), with guardrails. Turn off training/chat history where the tool allows, never paste customer, applicant, or confidential data, and verify outputs before using them. The vendors' own policies are the reason for the limits: most consumer tiers train on chats by default, and some retain human-reviewed conversations for years. Anything beyond general internal work, meaning customer data or confidential business information, needs a company enterprise seat (Claude, or Microsoft 365 Copilot for those who have it).

Official links
How OpenAI uses consumer dataHow Google uses consumer Gemini dataHow Anthropic uses consumer data

How to read these approvals

Approvals are tier-specific: each card states what is approved at which data tier today. The committee reviews and updates these on an ongoing basis, so always confirm here before using a tool for work. Enterprise seats (Claude, and Microsoft 365 Copilot for some teams) are limited, so request a seat if your work needs one.