AI and Technology Claims in Marketing

Under the SEC Marketing Rule, any marketing claim that you “use AI,” “machine learning,” or other advanced technology must be accurate, substantiated, and not misleading. Operationalizing this means inventorying every AI/tech claim in advertisements, tying each claim to written substantiation, and routing all future AI/tech language through a compliance review workflow aligned to 17 CFR § 275.206(4)-1(a). (17 CFR § 275.206(4)-1)

Key takeaways:

• Treat “AI” and “technology advantage” statements as testable claims that require evidence, not branding language. (17 CFR § 275.206(4)-1)
• Build a substantiation file per claim: what is used, where, by whom, with what limits, and since when. (17 CFR § 275.206(4)-1)
• Lock in a repeatable pre-approval process so marketing cannot publish new AI claims without Compliance sign-off and retained backup. (17 CFR § 275.206(4)-1)

“AI and technology claims in marketing” is a plain-language way to describe a recurring exam and enforcement risk: firms overstate what their tools do, imply capabilities they do not have, or blur the line between experimentation and production use. Under the SEC Marketing Rule’s general prohibitions, an advertisement cannot include any untrue statement of material fact, or omit a material fact necessary to make the statement not misleading. AI and “data-driven” language is not exempt; it is evaluated like any other performance, process, or capability claim. (17 CFR § 275.206(4)-1)

For a CCO or GRC lead, the operational problem is predictable. Marketing content moves fast, AI terminology is fuzzy, and multiple teams contribute language (portfolio, quant, product, IR, third parties). If you do not create a single source of truth for what “AI” means at your firm, you will end up with inconsistent disclosures, overbroad claims, and weak substantiation. This page gives you a requirement-level playbook: scope the requirement, define what “substantiation” looks like, implement a review workflow, and retain the evidence an examiner will ask for. (17 CFR § 275.206(4)-1)

Regulatory text

Requirement (operator-ready): Marketing claims about the use of artificial intelligence, machine learning, or other technology must be substantiated and not misleading, consistent with the SEC Marketing Rule’s general anti-fraud prohibitions. (17 CFR § 275.206(4)-1)

What you must do in practice:

1. Identify AI/tech statements in any “advertisement” and treat them as factual claims. (17 CFR § 275.206(4)-1)
2. For each claim, document support showing the statement is true, complete, and not presented in a way that creates a misleading impression. (17 CFR § 275.206(4)-1)
3. Implement pre-approval and recordkeeping so the firm can prove, after the fact, why the claim was reasonable when published. (17 CFR § 275.206(4)-1)

Plain-English interpretation (what the SEC expects you to get right)

If your website, pitch deck, RFP response, fact sheet, social post, or investor letter says you “use AI” or have a “proprietary machine-learning model,” you need to be able to back that up with concrete, dated evidence. “Back it up” means more than a slide from the data science team. It means: the tool exists, is actually used in the investment or operational process described, works in the way implied, and has limits disclosed when needed to avoid misleading impressions. (17 CFR § 275.206(4)-1)

Common claim types that require substantiation:

• Capability claims: “AI-powered portfolio construction,” “machine learning identifies mispricings.” (17 CFR § 275.206(4)-1)
• Differentiation claims: “Our AI gives us an edge,” “technology-driven risk management.” (17 CFR § 275.206(4)-1)
• Process claims: “We systematically ingest alternative data,” “automated model governance.” (17 CFR § 275.206(4)-1)
• Third-party tool claims: “We use [provider] AI,” “institutional-grade NLP.” You are still responsible for accuracy. (17 CFR § 275.206(4)-1)

Who it applies to (entity + operational context)

In-scope entities: SEC-registered investment advisers and fund managers preparing or distributing advertisements. (17 CFR § 275.206(4)-1)

In-scope activities: Any communication that meets the Marketing Rule’s concept of an “advertisement,” including traditional marketing materials and many one-to-many communications and endorsement/testimonial contexts where applicable. Your risk increases when materials are distributed broadly, reused across channels, or repackaged by placement agents or other third parties. (17 CFR § 275.206(4)-1)

Teams you must involve:

• Marketing/IR (content owners)
• Portfolio management/research (process truth)
• Data/engineering (what is actually deployed)
• Compliance (approval + recordkeeping)
• Third-party management/procurement (tool due diligence, contracts, permitted uses) (17 CFR § 275.206(4)-1)

What you actually need to do (step-by-step)

1) Build a complete inventory of AI/tech claims

Create a single register of every AI/tech claim across channels:

• Website pages, blogs, SEO landing pages
• Pitch decks, DDQs/RFPs, tear sheets
• Investor letters, factsheets, presentations
• Social posts and videos
• Third-party distributed material (placement agents, affiliates) (17 CFR § 275.206(4)-1)

Operational tip: Inventory “near claims” too: “quantitative,” “systematic,” “algorithmic,” “data-driven,” “automation,” “proprietary models.” These often imply AI even if the word is absent. (17 CFR § 275.206(4)-1)

2) Normalize each claim into a testable statement

For each claim, rewrite it internally as:

• What exactly is being claimed?
• Where is it used (investment process step or business function)?
• Is it in production, pilot, or roadmap?
• Is the output determinative, advisory, or purely administrative? (17 CFR § 275.206(4)-1)

Example normalization:

• Marketing line: “AI-driven security selection.”
• Testable statement: “A model that meets our internal definition of AI is used in live security selection decisions for Strategy X, under documented oversight and constraints.” (17 CFR § 275.206(4)-1)

3) Create a substantiation file per claim (minimum viable evidence)

For each claim, collect and retain:

• Process mapping showing where the tool fits (inputs, outputs, decisions affected)
• System evidence (architecture diagram, model card or technical summary, change logs)
• Governance evidence (approval to deploy, model review minutes, monitoring notes)
• User evidence (SOPs, screenshots, access logs, training materials)
• Scope boundaries (what strategies/accounts use it; what is excluded)
• Third-party documentation when a third party provides the AI component (contract scope, product description, internal validation notes) (17 CFR § 275.206(4)-1)

Your goal is not perfection. Your goal is defensibility: a clear chain from claim → reality → controls → limitations. (17 CFR § 275.206(4)-1)

4) Add “AI claim gates” to the marketing review workflow

Update your marketing review checklist to include:

• Does the content contain AI/ML/algorithmic/data-driven language? (Y/N)
• If yes, is there a claim ID tied to an approved substantiation file? (Y/N)
• Does the content imply outcomes or advantages that are not supported? (Y/N)
• Are limitations needed to prevent misleading impressions? (Y/N)
• Has Compliance approved the final layout, not just the draft text? (Y/N) (17 CFR § 275.206(4)-1)

Practical control: Require marketing to reference the claim ID in the submission ticket. No claim ID, no approval. (17 CFR § 275.206(4)-1)

5) Fix language that is “true but still misleading”

Common patterns to tighten:

• Overbreadth: “We use AI across portfolios” when only one sleeve uses a model. Narrow the scope. (17 CFR § 275.206(4)-1)
• Implied autonomy: “AI manages risk” when it produces analytics reviewed by humans. Describe human oversight. (17 CFR § 275.206(4)-1)
• Implied uniqueness: “proprietary AI” when it is mostly a third-party tool with light configuration. Be specific about what is proprietary. (17 CFR § 275.206(4)-1)

6) Extend controls to third parties and affiliates distributing your materials

If a placement agent, platform, or affiliate markets on your behalf:

• Provide an approved content pack with locked language for AI/tech claims.
• Prohibit edits to AI/tech sections without Compliance approval.
• Require periodic attestations that only approved materials are in use.
• Retain the third party’s distributed versions as part of your records. (17 CFR § 275.206(4)-1)

7) Operationalize recordkeeping for exams

Maintain a “marketing substantiation binder” that is easy to produce:

• Claim register (current + retired)
• Substantiation files
• Approval tickets and sign-offs
• Final materials as distributed (PDFs, web snapshots, social archives) (17 CFR § 275.206(4)-1)

Daydream can reduce friction here by centralizing claim registers, mapping claims to evidence, and keeping approvals and artifacts tied to the exact version that shipped, which is what exam teams test in practice. (17 CFR § 275.206(4)-1)

Required evidence and artifacts to retain (exam-ready list)

Minimum set you should be able to produce on request:

• AI/tech claim inventory and ownership (who approves, who maintains) (17 CFR § 275.206(4)-1)
• Substantiation file per claim (as described above) (17 CFR § 275.206(4)-1)
• Marketing review policy/procedure and review checklist with AI claim gate (17 CFR § 275.206(4)-1)
• Proof of pre-use approval for each advertisement version (ticketing exports, emails, or system audit trails) (17 CFR § 275.206(4)-1)
• Copies of advertisements as distributed, plus dates/channels of distribution (17 CFR § 275.206(4)-1)
• Third-party distribution controls and attestations (if applicable) (17 CFR § 275.206(4)-1)

Common exam/audit questions and hangups

Expect questions like:

• “Show me all places you claim to use AI or machine learning.” (17 CFR § 275.206(4)-1)
• “What do you mean by AI at this firm?” (17 CFR § 275.206(4)-1)
• “Which strategies actually use these models in production?” (17 CFR § 275.206(4)-1)
• “Who reviewed and approved this claim before use?” (17 CFR § 275.206(4)-1)
• “What evidence did you rely on to conclude the claim is not misleading?” (17 CFR § 275.206(4)-1)
• “Do third parties market your AI capabilities, and how do you supervise them?” (17 CFR § 275.206(4)-1)

Hangups that slow teams down:

• No single owner for “what is true” about technology.
• Engineers describe prototypes; marketing treats them as deployed.
• Substantiation exists but is scattered and not tied to specific language versions. (17 CFR § 275.206(4)-1)

Frequent implementation mistakes (and how to avoid them)

1. Defining AI too loosely.
   Fix: Create an internal definition used for marketing claims and require mapping to systems or workflows that meet it. (17 CFR § 275.206(4)-1)

2. Relying on intent instead of evidence.
   Fix: No claim without a substantiation file that includes proof of actual use, not planned use. (17 CFR § 275.206(4)-1)

3. Leaving “edge” claims unbounded.
   Fix: Replace “gives us an edge” with specific, supportable descriptions of process improvements, and add limitations when needed to avoid misleading impressions. (17 CFR § 275.206(4)-1)

4. Assuming a third party’s marketing is safe to repeat.
   Fix: Treat third-party product descriptions as inputs, then validate what your firm actually uses and how. (17 CFR § 275.206(4)-1)

Enforcement context and risk implications

The SEC has indicated that misleading “AI” and technology claims can be evaluated under the Marketing Rule’s general prohibitions, and the staff has highlighted “AI washing” risk in the market. Your practical exposure is not limited to intentionally false statements; it includes overbroad wording, missing qualifiers, and outdated claims that remained published after the underlying process changed. (17 CFR § 275.206(4)-1)

Practical 30/60/90-day execution plan

First 30 days: Stabilize and stop new risk

• Freeze new AI/tech marketing language unless Compliance approves against a substantiation file. (17 CFR § 275.206(4)-1)
• Inventory existing materials and flag high-risk claims (broad, outcome-implying, or hard to evidence). (17 CFR § 275.206(4)-1)
• Assign accountable owners: Compliance (approval), Marketing (content), Tech/Quant (substantiation). (17 CFR § 275.206(4)-1)

Next 60 days: Build the substantiation system

• Normalize each claim into a testable internal statement and decide whether to keep, narrow, or retire it. (17 CFR § 275.206(4)-1)
• Create substantiation files for retained claims and link them to the exact language used. (17 CFR § 275.206(4)-1)
• Update marketing review procedures and checklists to include AI claim gates and third-party distribution controls. (17 CFR § 275.206(4)-1)

Next 90 days: Make it durable

• Train Marketing/IR and portfolio teams on “claim discipline” and how to request approvals. (17 CFR § 275.206(4)-1)
• Implement periodic re-validation when models, data sources, or workflows change so claims do not go stale. (17 CFR § 275.206(4)-1)
• Centralize records (claim register, evidence, approvals, final artifacts). A platform like Daydream can serve as the system of record for claims and substantiation across channels. (17 CFR § 275.206(4)-1)

Frequently Asked Questions

Do we have to stop saying “AI” entirely?

No. You need to ensure the statement is accurate, can be substantiated, and is not presented in a misleading way under the Marketing Rule. If you cannot substantiate it, narrow the claim or remove it. (17 CFR § 275.206(4)-1)

What counts as “substantiation” for an AI claim?

Substantiation is documentation that demonstrates the claim is true in the way a reasonable investor would understand it, plus enough context to avoid misleading impressions. In practice, that includes process maps, governance evidence, and proof of real use in the described workflow. (17 CFR § 275.206(4)-1)

Our model exists, but it’s only used in research, not trading. Can we still say “AI-driven investing”?

That phrasing can imply the model drives investment decisions. If the model only informs research, describe it that way and include human oversight and scope boundaries so the ad is not misleading. (17 CFR § 275.206(4)-1)

Can we rely on a third-party AI provider’s documentation as support?

You can use it as supporting material, but you still need evidence of how your firm actually uses the tool and what the tool does in your environment. Your claim must match your implementation, not the provider’s broad marketing. (17 CFR § 275.206(4)-1)

What if different strategies use different levels of automation?

Avoid firmwide statements unless they are true for all covered products. Use strategy-level language and maintain separate substantiation where the process differs. (17 CFR § 275.206(4)-1)

How do we handle legacy pitch decks already in the field?

Treat them as live risk if they are still being shared. Pull them back into the review workflow, update the AI/tech language to match current reality, and retain the prior versions and remediation notes for your records. (17 CFR § 275.206(4)-1)