Claremont SecurityAI operations · Compliance
Advisory and compliance

Services

Claremont Security serves firms entrusted with sensitive client data across the United States: a fixed-fee two-week review, then AI set up on the tools your team already uses, the written security plan and compliance work your rules require, and ongoing care.

More work done with AI, without putting client data at risk.

Which rules apply depends on the kind of business. The industries page sets them out for accounting, tax and law practices and the other sectors it covers. Who we work with →

Four steps. Every fee fixed and in writing before work starts.

  1. 01Estimate

    Your numbers on one page

    What one improvement is worth to your firm, what it takes, and how soon it pays back.

    Fee none

  2. 02Assessment

    Baseline Security Assessment

    Two weeks, read-only. Every gap, and which AI tools you can use on client work today.

    Fixed fee · credited in full against the engagement that follows, if it starts within 90 days

  3. 03Rollout

    AI rollout, Copilot or Claude

    Gaps closed highest risk first, file access fixed, the tool set up, and staff trained.

    Fee fixed per engagement · engagements and fees

  4. 04Ongoing care

    Kept current every quarter

    As your staff, the IRS forms and the tools change, with questionnaires answered and a quarterly review.

    Fee monthly · cancellable at renewal

Scope

The three categories

Who each one is for, what it needs from you, and what you get. Fixed fee wherever the scope is fixed.

// Category_01

AI in daily operations

Staff get AI in a setup you control.Our position for every tool: until the vendor commits to U.S.-only processing and the firm can show who reaches the data, client tax return information goes in only with each client's consent.

Built forFirms that want more work done with the same staff: rolling out Copilot, already on a Claude or ChatGPT business plan, or just finding out staff use personal AI on client work.
RequiresFor a Copilot rollout: Microsoft 365 Business Premium plus a Copilot license for each person who uses it. For Claude or ChatGPT: the firm’s business plan. Personal plans are out of scope.
OutcomeStaff using AI every day on real work, under a business agreement your firm signed. On Copilot or Claude Enterprise, a record of who used it for what.
IPermission review before rollout
  • Copilot shows a person exactly what that person could already open, so the permission work comes first, never after
  • SharePoint and OneDrive checked against who should have what, before Copilot is switched on
  • It is rarely the AI. It is the folder somebody shared with the whole firm three years ago and forgot
What you walk away with

A list of everything that is over-shared, named by site and library.

IITenant boundary and data loss prevention
  • Microsoft commits to storing prompts and responses in the United States for U.S. tenants, not yet to running the AI only there
  • Prompts and responses are not used to train the underlying models
  • Staff see the warning while they are still typing, not in a report a month later
What you walk away with

Labels and policies switched on, the vendor’s no-training terms checked and filed, and one page saying where your data can and cannot go.

IIIGoverned agents
  • Help with daily drafting for everyone, the morning inbox in Copilot, and one tool for the work your practice does most
  • Fixed output standards and a person's review. Nothing any of them writes sends itself
  • Anything that has to go outside the tenant goes through the Clean Room, which is free
What you walk away with

Each one set up for your firm, AI use logged where the plan carries it (Purview on Copilot, the Compliance API on Claude Enterprise), and one page per agent saying what it may do and what it must refuse.

Diagram of a firm's Microsoft 365 tenant and its boundary. A managed staff workstation's request to Microsoft 365 Copilot is allowed and logged. Its paste of client identifiers toward a free AI chatbot is blocked at the boundary by a data loss policy, beside sensitivity labels and Conditional Access. Outside the boundary sit personal webmail, a file-sharing site, a public AI chatbot and an unmanaged laptop. A record below lists both requests.Where client data can go, and where it can'tOpen the diagram →
// Category_02

AI Oversight and Virtual CISO

Too small to justify a security hire.Too exposed to go without one. We lock down who can sign in, check what your existing tools actually cover, and answer for your security in writing.

Built forFirms with nobody whose job is security, and a growing pile of client security questionnaires.
RequiresMicrosoft 365 Business Premium. Business Standard has neither Intune nor Defender for Business, so this cannot be delivered on it.
OutcomeA defended firm with coverage you can point at, and one advisor who answers the questionnaires, sits down with your leadership once a quarter, and picks up the phone when something looks wrong.
IIdentity and access control
  • Multi-factor authentication on every account, the admin and service accounts included
  • Those are the accounts that usually get an exemption, and the exempted group is where the real gap lives
  • Offboarding runs from a written procedure rather than from memory
What you walk away with

Sign-in rules switched on with every exemption written down, a register of who holds admin rights, and an offboarding checklist that cuts access on day one.

IIEndpoint coverage verification
  • On Business Premium you already pay for Defender. The work is proving it reaches every machine
  • The problem is rarely a missed detection. It is the three laptops nobody knew were unmanaged
  • Update status confirmed at the machine, not read off a dashboard
What you walk away with

Defender set up to a written standard, coverage checked against your staff list, and update status confirmed machine by machine.

IIIWhat maintenance adds
  • Your AI setup kept current as the tools change
  • One quarterly review with your leadership, with a short note on what changed in your AI tools and what to do about it. It is the only standing commitment.
What you walk away with

Every control at deployment rather than held back for a higher tier, and a setup that improves instead of one frozen on the day it was built.

// Category_03

Compliance Documentation

A firm can be secure and unable to prove it.That gap is the exposure. We turn what you already do into the record a regulator expects to read.

Built forTax and accounting firms under the FTC Safeguards Rule, which requires a written security plan. Especially the ones whose plan is a template nobody has opened.
OutcomeA written record of your security program that holds up when someone asks for it. Where an assessment is run, a list of gaps with owners and dates. The written plan itself is Engagement III, Compliance & WISP.
IBaseline Security Assessment
  • For firms that need proof, not just paperwork. Two weeks of looking from the day we get access
  • Nothing is touched. Fixed fee, credited in full against the engagement that follows, if it starts within 90 days
  • This is step 02 of the engagement above, not a second purchase
What you walk away with

A list of every gap against 16 CFR Part 314 and IRS Publication 4557, each with an owner by role and a date. Plus which AI tools your firm can use on client work today, and what each needs first.

II§7216 consent review
  • For firms using an AI tool, or sending returns to preparers outside the United States
  • We list what your client consent form must mention for each AI tool you use
  • The wording itself comes from your lawyer or your tax software
What you walk away with

What your client consent page must name for each provider, including anyone outside the United States who may receive or view return information, its support staff included. For Form 1040 clients the Social Security number stays out even with consent.

Limits

What we do not do

Stated plainly, because the gap between what a firm needs and what an outside advisor can deliver is exactly where engagements go wrong.

We do not monitor a console around the clock

The platform monitors continuously. We review and respond to alerts on a defined business-hours cadence. We do not run a help desk or an around-the-clock response team. For overnight cover you need a managed detection provider, and we'll say so.

We do not deploy patches

Category II is oversight and evidence. Known-exploited vulnerabilities are surfaced and tracked to closure with per-endpoint evidence, and your IT provider does the deploying.

We do not issue formal sign-offs

Formal opinions and sign-offs belong to licensed firms. Claremont performs assessments and writes documentation you can hand to whoever asks.

We do not run phishing simulations

No platform for it is in place. If that is a requirement, say so early and we will tell you who does it.

We do not take more than two engagements at once

Two simultaneous engagements is the honest ceiling. If the calendar is full you will be told, not queued quietly.

The rest of it

How the function is structured from outside, what it covers, the cadence it runs on and what it runs on. The fractional model brief →

Start with a one-page estimate, or run the AI Exposure Check on your own tools and send us what it shows.

Get a one-page estimate Talk to us Run the instrument

Security or compliance team at a larger organization, or a firm subcontracting work: enterprise and partner engagements.