Is It Safe to Upload Architecture Client Documents to AI? A Sort by Document Type

Is it safe to upload architecture client documents to AI? For a fee proposal on a business account, usually yes. For a security layout under NDA, no account makes it safe. The answer changes with the file, so sort the file first.

TL;DR: Two decisions settle it. The first is the account: business and API plans from the major providers exclude your content from training by default, and personal plans may need a setting changed. The second is the document: scope, fee and programme text sits in a green group that is fine to upload, files with a named client at a private address belong in an amber group that you strip first, and security details, NDA material and personal data about a client stay out of general AI tools. Once a document is cleared, DocsAura, an AI document design tool, takes the DOCX or PDF your studio already wrote and returns a designed page in about two minutes.

Is It Safe to Upload Architecture Client Documents to AI? Start With Two Decisions

Studios tend to ask this as one question about one tool. It is two questions, and each has a short answer.

Decision one is the account. OpenAI states that by default it does not use data from ChatGPT Enterprise, Business, Edu or its API platform for training or improving its models.[^1] Anthropic says the same about its commercial products, such as Claude for Work and the API.[^2] On Anthropic's personal plans (Free, Pro, Max), chats are used for model improvement if you choose to allow it, if a conversation is flagged for safety review, or if you opt in to a programme, and Incognito chats are excluded.[^3] A thumbs-up or thumbs-down press sends the whole conversation to the provider on both commercial and personal Claude plans, and Anthropic keeps feedback conversations for up to five years.[^2] A firm AI policy template published by Archipractica, which compared vendor pages on 20 October 2026, reads OpenAI's personal plans the same way as Anthropic's: content may be used for training unless the setting is switched off.[^4]

Decision two is the document. A business plan keeps your file out of training. It leaves untouched the question of what the file contains and what you promised the client about it. That promise sits in your appointment, in any NDA, and in the professional duty of confidentiality.

What Your Professional Duty Says About Client Information

The AIA Code of Ethics, Rule 3.401, says members shall not knowingly disclose information that would adversely affect their client or that they have been asked to maintain in confidence, except as the Code or law allows. The AIA Trust, writing with Victor Insurance Managers in February 2025, applies that rule to generative AI and gives two examples. A high-security facility such as a prison or a defence office is one. A residential client who values privacy is the other, where photos showing the entry and exit points of the home can affect the client's sense of security.[^5]

The same article lists what to put in a firm policy: which AI tools are permitted, what information may go into them, and training on how to scrub sensitive client, project and firm information before use.[^5] It also says to read each client contract for restrictions on AI.

The scale of the question is large. RIBA's 2026 survey reports that AI use among architecture practices rose from 59% in 2025 to 74% in 2026.[^6] Around one in five practices has a documented AI policy, according to Beale & Co's summary of the same report.[^7] Most studios are already uploading files, many of them with no written rule about which ones.

What we found reading three architecture AI-confidentiality guides

On 28 October 2026 we read three guides written for architects on protecting client information when using AI: the AIA Trust article on ethical challenges of generative AI, Archipractica's firm AI policy template, and Studio Matrx's client confidentiality lesson. All three tell the studio to check the client contract before choosing a tool, and all three recommend a written firm rule. Two of the three name the account plan as the setting that matters. All three say to scrub identifying details first, and only Studio Matrx lists the fields and tells you to crop the title block from a drawing. None sorts the studio's routine paperwork, such as a fee proposal or a stage report, by what is written inside it.

Sort Studio Paperwork Into Green, Amber and Red

The sort below uses what each document contains. The grouping is our own, built from the guidance above, and it assumes a business-grade account. Your contract and any NDA outrank it.

Two habits cover most of the amber group. Studio Matrx suggests replacing identity with description, so that a named family at a named plot with an exact budget becomes a family of four on an urban infill plot with an upper-middle budget.[^8] For drawings and PDFs, the same lesson says to crop the title block, the client name and any site plan carrying the address, because the tool reads what is on the sheet and the sheet is what you disclosed.[^8]

A Five-Minute Pass Before You Upload

1. Open the contract. Check for any clause on AI or on disclosing project information to third parties. Contract silence on AI gives you no permission, as Archipractica's template puts it, and the policy owner should be asked first.[^4]
2. Search the file for the client's name and the site address. Check the header, footer, file name and document properties as well as the body.
3. Band the money. A fee proposal needs your fee. The client's total construction budget can often be written as a range.
4. Look at comments and tracked changes. Internal notes about a client travel with the file.
5. Pick the account. Use the business seat or the API. If staff use a personal login for project files, switch training off and record the date, as the policy template advises.[^4]
6. Skip the feedback buttons. On Claude, a thumbs-up or thumbs-down sends the whole conversation to Anthropic, and Archipractica reports the same risk on ChatGPT personal plans.
7. Tell the client. The AIA Trust advises being upfront about GenAI use, so that anyone uncomfortable with it says so before the work starts.[^5]

Where DocsAura Fits

DocsAura, an AI document design tool, takes a DOCX or PDF the studio already wrote and returns a designed HTML page. It does not draw, model or generate plans, and it has no integration with Revit, ArchiCAD or AutoCAD. It works on one finished document at a time, so what goes in is the file you chose, after the pass above.

DocsAura's privacy policy lists an AI processing provider and hosting providers among those that receive the content you submit.[^9] Read it as you would any tool's terms, and treat a DocsAura upload by the same green, amber and red sort. A green document such as a fee proposal or a stage report needs nothing removed. An amber one goes in after the placeholder swap, and you can type the real name back in the editor once the page comes back. A red one stays with your own team.

That sort also shortens the trial. The designing is the part of a fee proposal or stage report that costs a studio an evening, and the green group holds most of a studio's weekly paperwork. Our guides on writing a site visit report clients understand and on the AI tools small practices use cover the documents themselves. For the redaction detail in any industry, see what to redact before uploading a document to AI.

Try It on One Green Document

Open the most recent green document you have: a fee proposal you already sent, last quarter's stage report, a scope of services. Check it against the contract and the five-minute pass, then drop that one file into DocsAura, the AI document design tool, and see what comes back in about two minutes. There is nothing to install and no template to build. If the page suits the studio, send it. If a section needs a change, edit it in the editor and export.

[^1]: OpenAI, "Business data privacy, security, and compliance," openai.com/business-data, read 28 October 2026. [^2]: Anthropic Privacy Center, "Is my data used for model training?" (commercial products), privacy.claude.com, 18 August 2026. [^3]: Anthropic Privacy Center, "Is my data used for model training?" (consumer products), privacy.claude.com, 16 March 2026. [^4]: Archipractica Editorial, "Architecture Firm AI Policy: Free Template + Rollout Guide," archipractica.ai, updated 20 October 2026. Vendor settings table checked by Archipractica on 20 October 2026. [^5]: AIA Trust and Victor Insurance Managers, "Ethical challenges of generative AI in architectural practice," theaiatrust.com, February 2025, quoting AIA Code of Ethics and Professional Conduct Rule 3.401. [^6]: Royal Institute of British Architects, "AI adoption reaches tipping point across UK architecture," riba.org, 22 July 2026. [^7]: Beale & Co, "RIBA AI Report 2026 charts the growing role of artificial intelligence in architectural practice," beale-law.com. [^8]: Studio Matrx, "Client Confidentiality & Data," Claude for Architects & Designers, Lesson 10.1, studiomatrx.org, September 2026. [^9]: DocsAura Privacy Policy, section 5 "Third-Party Data Sharing," docsaura.com/privacy, last updated 6 May 2026.