Feature, content, privacy, marketing, and launch rules for the United States and selected international markets
ClassicInvestor can be built as a research, calculation, and journaling product. It should not be launched as a paid, personalized investment decision service in its current full concept without written securities-law advice covering both the United States and Australia.
The decisive issue is product behavior, not whether the interface avoids the words "buy" and "sell." A paid service can move into regulated investment-advice territory when it applies analysis to a particular security or a user's portfolio, objectives, risk tolerance, time horizon, or saved thesis and then produces an opinion, prioritization, warning, score, ranking, suggested assumption, allocation implication, or action-oriented alert. Calling that output educational, a draft, a lens, or an AI response does not change its substance.
| Market | Current status | What may proceed | What is held |
|---|---|---|---|
| United States | YELLOW - conditional safe core only | Public-source facts, source-linked filing tools, user-authored journals, user-selected filters, and arithmetic using user-entered assumptions | Paid personalized chat, security opinions, portfolio-context outputs, allocation implications, app-generated scores/ranks, and action-oriented alerts pending written federal and state adviser-perimeter advice |
| Australia | RED - hard legal gate | Internal development and a tightly controlled factual prototype | Any Australian-facing or paid launch, and the proposed paid service generally, pending Australian counsel's AFSL/financial-product-advice opinion |
| United Kingdom | RED - geo-block until cleared | Internal testing with synthetic data | Consumer access pending FCA perimeter and financial-promotion advice |
| European Union / EEA | RED - geo-block until cleared | Internal testing with synthetic data | Consumer access pending MiFID II, GDPR, AI Act, market-abuse, and content/database review |
| Canada | RED - geo-block until cleared | Internal testing with synthetic data | Consumer access pending province-specific securities advice |
| Singapore | RED - geo-block until cleared | Internal testing with synthetic data | Consumer access pending Financial Advisers Act advice |
Ship, test, and market the product as a user-directed evidence workspace, not a decision engine. The product may make sources and the user's own reasoning easier to inspect. It must not decide what is attractive, suitable, urgent, underpriced, high quality, portfolio-appropriate, or worth acting on.
This assessment responds to every product, content, privacy, branding, beta, and international question in the source briefing. It is a product-perimeter assessment for build and launch planning. It is not a jurisdiction-specific legal opinion, a registration determination, or a substitute for retained U.S., Australian, UK, EU, Canadian, or Singapore counsel.
The assessment assumes:
Any change to those assumptions requires a new review.
The federal Investment Advisers Act generally reaches a person who, for compensation, is in the business of advising others about securities or issuing analyses or reports concerning securities. The statutory publisher exclusion is narrow: it protects a bona fide publication of general and regular circulation, not a personalized advisory relationship disguised as publishing.
The U.S. Supreme Court's decision in Lowe v. SEC is helpful to impersonal, disinterested publications offered generally and regularly. It is much less helpful where a service has personal interactions, learns a subscriber's circumstances, or tailors output to the subscriber's portfolio or needs. Automated delivery does not create a separate exemption; the SEC treats automated advice as a method of providing advisory services.
The U.S. position therefore depends on the full facts, including:
Required U.S. gate: before paid general availability or activation of personalized decision features, obtain a written opinion from U.S. investment-management counsel addressing federal adviser status, the publisher exclusion, state adviser registration, product marketing, and the exact feature set.
Australia defines financial product advice broadly as a recommendation or statement of opinion that is intended, or could reasonably be regarded as intended, to influence a decision about a financial product. Advice becomes personal where the provider has considered one or more of the person's objectives, financial situation, or needs, or a reasonable person might expect those circumstances to have been considered.
That line can be crossed without using an explicit command. A score, warning, valuation conclusion, portfolio-fit statement, personalized alert, or suggested allocation may be an opinion intended to influence. A general-advice warning is not a license and cannot cure unlicensed personal advice.
Required Australian gate: obtain written Australian financial-services counsel advice on whether the operator needs an Australian financial services licence, authorization under another licensee, a product redesign, or geographic/contract restrictions. This gate applies before a paid launch and before assuming a U.S.-only audience removes Australian exposure.
The UK, EU, Canada, and Singapore each distinguish information from regulated recommendation or advisory activity, but all can reach tailored, transaction-related, or inducement-oriented conduct. The current personalized feature set should not be exposed in those markets without local counsel and a market-specific implementation.
The following may be used to retrieve, organize, or navigate a user's own material without changing the substance of investment analysis:
The following must not be used to generate or change a security-specific opinion, score, priority, warning, recommendation, or action:
For a given source record and factual question, the factual answer should be materially the same for all users. User context may determine which saved document is opened or which user-created assumption is referenced; it must not determine whether the system says a security is attractive, risky, suitable, urgent, or worthy of action.
The system must refuse not only direct requests such as "Should I buy this?" but also multi-turn attempts to obtain the same conclusion indirectly. Refusal state must carry across the conversation. Examples include:
The refusal should redirect to source facts, user-entered criteria, or a blank worksheet without suggesting the answer.
The labels below mean:
| # | Feature | U.S. ruling | Required design | Prohibited implementation |
|---|---|---|---|---|
| 1 | Stock screener | CONDITIONAL | User chooses filters and thresholds; defaults are blank or purely mechanical; metrics are source-labelled; sorting is user-controlled | App-created "quality," "value," "best," or "moat" screens applied to live securities; default ranked results; app-selected thresholds presented as sound investing criteria |
| 2 | Process worksheets and valuation arithmetic | ALLOW / CONDITIONAL | User selects the model and enters every material assumption; output is arithmetic with sensitivity ranges; no conclusion | App-suggested assumptions, growth rates, discount rates, margins, target prices, "fair value," "undervalued," "overvalued," or buy/sell implication |
| 3 | AI first-pass evidence review | CONDITIONAL | Call outputs "source observations," "possible gaps," and "questions"; cite exact source spans; user may accept/edit; no user portfolio context | App-authored strengths/weaknesses as verdicts; positive/negative score; unverified paraphrase; treating user acceptance as legal cure |
| 4 | AI chat with memory | HOLD for personalized decision use | May retrieve cited facts and the user's own text; may explain a user-selected calculation; refusal is conversation-wide | Using portfolio, goals, horizon, tolerance, or history to opine on a security, prioritize opportunities, infer fit, suggest assumptions, or propose action |
| 5 | Thesis journal and decision record | ALLOW | User-authored; version history; clear edit and deletion behavior; export; timestamps | Calling a record legally immutable, signed, certified, or tamper-proof unless technically and legally true; retaining it contrary to deletion policy |
| 6 | Monitoring alerts | CONDITIONAL | User creates the exact monitor; event match is mechanical; source and date shown; neutral notification | "Thesis broken," "risk rising," "review now," urgency, red warning, action pressure, app-selected monitor, or ranked alert importance |
| 7 | Read-only portfolio sync | HOLD for normative output; CONDITIONAL for arithmetic | Display holding percentage and difference from a user-entered threshold; neutral typography; explicit consent and privacy controls | Default threshold; ideal allocation; target; over/underweight; trim/add; red/green judgment; suggested size; ranking holdings by need for action |
| 8 | Calibration and expected-v-actual | CONDITIONAL | Historical descriptive statistics about the user's recorded estimates; transparent sample size and method | Investor score, quality grade, "you are overconfident," future parameter suggestion, recommended behavioral correction, or security-selection output |
| 9 | 13F manager holdings | CONDITIONAL | Objective published manager-selection method; alphabetical or user-selected sort; report date and limitations; separate research panel | Popularity score; "smart money" label; manager-count signal; alerts based on buys/sells; placement beside the user's decision as social proof |
| 10 | Failure-pattern base rates | CONDITIONAL | Methodology, sample, period, definitions, uncertainty, missingness, survivorship and look-ahead limitations; same display for all users | Traffic-light conclusion, company failure probability, warning label, action implication, personalized presentation, or unsupported causal claim |
| 11 | Executive commitment ledger | PROHIBIT "broken promise" form | Quote exact public statement in context; neutral outcome status; objective criteria; human review; correction path | "Broken," "lied," "dishonest," "untrustworthy," credibility score, intent inference, cross-company blacklist, or automated adverse label |
| 12 | Enforcement and litigation | CONDITIONAL | Public source; procedural posture; allegation vs finding; date; updates; sealing/withdrawal logic; human review for named people | Guilt inference, sensational label, stale charge shown as final, sealed/expunged content, unnecessary personal data, or automated character judgment |
| 13 | Filing diffs and comment threads | ALLOW / CONDITIONAL | Identify exact compared documents and dates; label automated diff; cite changed language; show limitations | "Concealed," "hidden," "admitted," "red flag," or intent/importance inference unless the source directly establishes it |
User-owned filtering is the lowest-risk form. The product may let a user say, for example, "revenue growth greater than 10%," then show the companies that mechanically satisfy that condition. The interface must make clear that the user created the condition.
Templates create more risk because they express the publisher's judgment about what criteria matter. If templates are retained before counsel clearance, they should be educational blank structures, not live screens. A template may explain what a metric measures and ask the user to enter a threshold. It should not arrive with app-selected thresholds and immediately rank named securities.
Default sorting should be alphabetical, chronological, or the user's prior explicit choice. Do not default to a composite quality score or a metric that functions as a recommendation.
Arithmetic is safer than opinion. Every material assumption must be user-entered or copied from a source at the user's direction. If a source value is offered, it must be identified as a historical source value, not a recommended forecast.
Sensitivity tables may show what the formula produces across user-chosen scenarios. The product must not label a scenario "base," "reasonable," "conservative," or "most likely" unless the user created that label. Do not compare calculated values to market price using recommendation-like labels.
Replace evaluative headings such as "Strengths" and "Weaknesses" with source-oriented headings:
The system may identify tension between two cited statements. It should not decide whether the tension makes the security attractive or unattractive. Every material proposition must carry a source or the label "AI inference - verify."
Chat is the highest-risk surface because a conversational system creates a reasonable expectation that it is considering the user's circumstances. A banner alone is insufficient.
Until written counsel clearance, chat may:
It may not:
The alert must be a source-match notification, not an investment warning. Safe example:
New filing contains an item matching your saved monitor for Assumption A3. Source: [filing], filed [date]. No investment action is suggested.
Do not use red/green status, exclamation icons, countdowns, push-notification urgency, "thesis risk," or "material concern" unless the user wrote and applied that label to their own rule.
If a user has entered an 8% threshold and a synced position is 12%, the product may display:
Current position: 12%. Your saved threshold: 8%. Arithmetic difference: +4 percentage points.
It must not say "overweight," "above ideal," "reduce by 4%," "trim," or "risk is elevated." The threshold must come from the user; the product must not supply a default.
Descriptive feedback may say:
Across 9 completed records, 7 stored revenue estimates exceeded the later reported value. Median difference: 6%. This is a historical description, not a forecast or recommendation.
It must not say:
You are an optimistic investor. Lower future estimates by 6%.
13F filings are delayed and incomplete. They generally do not show a manager's complete current portfolio, shorts, many derivatives, or investment rationale. Product design must prevent manager-following from becoming a signal. Do not aggregate "number of respected managers holding" beside a decision memo or security card.
The proposed "kept/broken" label creates defamation, false-light, privacy, and consumer-deception risk. The product may report source facts, but should not infer dishonesty or intent. Use neutral, evidence-bounded statuses such as:
Every adverse status should receive human editorial review, contextual source quotation, publication date, objective criteria, a correction channel, and prompt update logic.
The current build should default to geo-blocking outside the approved U.S. beta. The entries below are not launch approvals; they identify what a local-counsel-cleared market version would likely need.
| Feature group | UK | EU / EEA | Canada | Australia | Singapore |
|---|---|---|---|---|---|
| Raw public facts, source links, filing diffs | Potentially permissible if factual and not a promotion/recommendation | Potentially permissible with GDPR, database, AI and market-abuse controls | Potentially permissible if non-tailored | Potentially permissible as factual information | Potentially permissible as factual information |
| User-authored journal | Generally lower risk | Generally lower advice risk; GDPR still applies | Generally lower risk | Generally lower risk | Generally lower risk |
| User-selected arithmetic | Conditional; no suitability or action implication | Conditional; no personal recommendation | Conditional; no tailoring into recommendation | Conditional; broad opinion/influence test | Conditional; no recommendation/inducement |
| Screener/templates | Hold templates that express a view | Hold ranking or recommendation-like screens | Hold tailored screens | Hold app-selected criteria/thresholds | Hold app-selected recommendation signals |
| AI security analysis | RED pending local counsel | RED pending local counsel | RED pending provincial counsel | RED pending AFSL advice | RED pending FAA advice |
| Personalized chat | RED | RED | RED | RED | RED |
| Portfolio comparison/allocation | RED | RED | RED | RED | RED |
| Action-oriented alerts | RED | RED | RED | RED | RED |
| 13F/social-proof signals | Conditional facts only; financial-promotion risk | Conditional facts only; investment-recommendation risk | Conditional facts only | Conditional facts only | Conditional facts only |
| Named-person adverse labels | Hold pending defamation/privacy review | Hold pending defamation/privacy/GDPR review | Hold pending defamation/privacy review | Hold pending defamation review | Hold pending defamation review |
The UK perimeter includes regulated advice and the separate financial-promotion regime. Software-generated signals can amount to advice, and putting material on the internet does not by itself make it an impersonal publication. A disclaimer is ineffective if the service says or implies that an investment suits the customer or if the interaction gives a reasonable expectation that the customer's circumstances were considered.
Any invitation or inducement to engage in investment activity may also be a financial promotion. Consumer-facing marketing, social posts, app-store copy, influencer content, and in-product prompts must be reviewed, not only the recommendation engine.
UK gate: written FCA perimeter and financial-promotion advice, including whether the service requires authorization, an authorized approver, or a strictly non-personal factual version. UK GDPR work must be complete before targeting UK users.
MiFID II defines investment advice around a personal recommendation concerning transactions in financial instruments. Personalized security, portfolio, or action outputs are therefore high risk. Broadly distributed public commentary may also create investment-recommendation obligations under market-abuse rules if it suggests an investment strategy or expresses an opinion about value or price.
The EU version would also require GDPR territorial-scope and transfer controls, AI transparency, data-protection impact analysis where warranted, and database/content rights review. The AI Act requires transparency for interactions with AI systems in relevant circumstances, but an "AI" label does not solve securities regulation.
EU/EEA gate: local MiFID II and market-abuse advice, GDPR readiness, EU representative analysis, lawful transfer mechanism, AI transparency, and content/database rights.
Canadian securities registration is province- and territory-based under the national instrument framework. General, non-tailored commentary may be treated differently from advice tailored to a recipient. The present use of portfolios, goals, tolerances, saved theses, and conversational context is not a safe general publication posture.
Canada gate: written advice for the provinces to be targeted, with Ontario, British Columbia, Alberta, and Quebec analyzed before launch. Do not assume a U.S. disclaimer or publisher theory carries across.
The broad statutory test captures recommendations and opinions intended, or reasonably regarded as intended, to influence a financial-product decision. Personal circumstances substantially increase risk. The proposed paid personalized experience is not approved for Australian users and may affect the provider's status even where users are overseas.
Australia gate: written AFSL/perimeter advice on the exact product, provider entity, user geography, marketing, and data flow. A statutory general-advice warning is relevant only within an authorized general-advice model; it is not permission to provide unlicensed advice.
Singapore's Financial Advisers Act requires licensing for regulated financial-advisory services, subject to exemptions, and can deem overseas conduct to occur in Singapore where it is intended or likely to induce the Singapore public to use advisory services. Personalized recommendations and targeting create the greatest risk.
Singapore gate: written FAA scope and exemption advice before any targeting, onboarding, payment, or access.
Where a market is held, use more than an IP notice. Controls should include:
Geo-blocking is a risk control, not a guaranteed jurisdictional shield.
The source country is not a legal switch. Analyze at least four separate dimensions:
Consequently:
U.S. government-created SEC material may generally be copied subject to source, logo, endorsement, and access rules. Issuer filings and third-party exhibits require separate treatment because filing them with the SEC does not automatically place all third-party content in the public domain.
Allowed default:
Not allowed by default:
Court opinions, docket facts, and publicly available filings may be reported with careful attribution and procedural context. Party filings and exhibits may contain copyrighted, sealed, confidential, expunged, defamatory, or highly sensitive material.
Required controls:
Use short, attributed excerpts with a link to the original source. Prefer an EDGAR equivalent when available. There is no universal word-count safe harbor. Systematic ingestion of full reports, presentation decks, images, videos, or website archives requires permission or a documented fair-use analysis.
Kill by default unless licensed. Do not ingest, store, quote, summarize for users, embed, or use unlicensed commercial transcripts as LLM input merely because they can be found online.
A written license must expressly cover the intended activities, including:
A binary signals-only / display flag is not sufficient. Every source must have a rights record that answers:
The application must fail closed when a requested use is not affirmatively allowed by that record.
Treat portfolio holdings, account identifiers, balances, saved theses, loss tolerance, goals, financial reasoning, behavioral calibration, and investment history as high-sensitivity data. Some categories may not be "special category" data under GDPR, but the practical harm from misuse, exposure, profiling, or unauthorized advice is substantial.
Before live portfolio sync:
Default rule:
ClassicInvestor does not use journal, portfolio, chat, or decision-record content to train shared models or improve other users' outputs.
Any later use must have a separate, specific, optional, unbundled consent. Refusal must not reduce core service quality. Withdrawal must stop future use. De-identification must be irreversible in practice, documented, tested, access-restricted, and aggregated above a minimum cohort threshold.
Do not promise that an AI provider "never retains" or "never trains" unless the executed agreement, account settings, abuse-monitoring rules, and subprocessor terms support that exact statement.
Build the following as baseline product rights rather than waiting for a statutory threshold:
Recommended operational baseline:
The published policy must describe exceptions accurately. It must not claim immediate erasure from immutable backups if that is not technically true.
The product needs a U.S. privacy analysis covering state consumer-privacy thresholds and financial-data rules. Paid investment-related services and receipt of customer information from financial institutions also require a specific GLBA/Safeguards Rule analysis. Until counsel resolves applicability, build to the Safeguards Rule security baseline rather than relying on non-coverage.
Required operational controls include a written security program, risk assessment, access controls, encryption, service-provider oversight, incident response, secure development, logging, testing, and breach-notification mapping. Every U.S. state and relevant territory has a breach-notification regime.
Before targeting EU/EEA or UK users, complete:
Do not use absolute copy such as:
Each statement needs an auditable technical and contractual basis.
ClassicInvestor is not cleared for public expansion. No exact public-register hit is not the same as clearance. The relevant field already contains prominent MOAT uses, including financial-services and investment-research uses. VanEck has a U.S. registration for MOAT in financial services, and Morningstar identifies Moat Ratings among its trademarks. Adding "Brain" may not avoid confusion where the product itself provides investment research and decision tools.
Choose one of two paths before public marketing, filing, paid acquisition, app-store submission, press, or material brand spend:
Domain ownership, company-name availability, social handles, and an AI-generated search are not trademark clearance.
Disclosures must be conspicuous, readable, and placed at the relevant decision point. They must not be hidden behind a tooltip or terms link. They supplement the design restrictions; they do not replace them.
ClassicInvestor is a research, calculation, and journaling tool. It does not provide investment, legal, tax, or accounting advice and does not recommend any security, transaction, portfolio allocation, or course of action. You make and remain responsible for every investment decision. Investing involves risk, including loss of principal. Information may be incomplete, delayed, or inaccurate; verify material information in the cited source.
Replace the product name if the rebrand proceeds.
AI-generated research draft. Not a recommendation. Verify it against the cited source. It does not become part of your thesis unless you accept or edit it.
Source: [source]. As of [date/time]. Missing fields are shown as missing. This information may have changed.
This calculation applies the model and assumptions you entered. ClassicInvestor does not select those assumptions or determine fair value.
Arithmetic comparison to a threshold you entered. ClassicInvestor does not determine an appropriate allocation or recommend that you buy, sell, hold, add, or reduce.
New source item matched your saved monitor. No investment action is suggested.
13F data reports certain long U.S. holdings as of [date] and may be stale or incomplete. It does not show current holdings, short positions, or a manager's full portfolio, and is not an endorsement.
Status reflects cited public evidence under stated criteria as of [date]. It is not a finding of dishonesty or intent. Submit corrections at [link].
Private beta; features may change or contain errors. Do not rely on the service as the sole basis for an investment decision.
If and only if Australian counsel confirms an authorized general-advice model, use the statutory form counsel approves. A typical warning begins:
This information is general and has not been prepared taking into account your objectives, financial situation or needs. Before acting, consider whether it is appropriate for you and, where relevant, read the applicable Product Disclosure Statement.
Do not display this warning as a purported cure for unlicensed personal or general advice.
The prohibition applies to AI output, labels, notifications, templates, empty states, marketing, testimonials, metadata, structured events, emails, and generated documents.
Any outcome or AI claim requires reliable substantiation and must still remain inside the advice perimeter.
Every material factual output should include:
Uncited AI claims must be blocked or labelled "AI inference - verify," and must never support an adverse named-person label or a recommendation-like conclusion.
When data is missing, stale, contradictory, unlicensed, or not verifiable, the product must fail closed:
Log enough to test compliance without retaining unnecessary financial data. Required test categories include:
A private beta should be treated as a controlled product test, not a legal exemption.
Do not treat beta participation as consent to use a person's name, quote, company, image, portfolio, or performance story in marketing. Obtain a separate, specific release. Disclose any free access or other material connection. Do not condition access, compensation, or continued participation on positive feedback.
The following must not ship without a later written legal approval tied to the exact implementation, and several should not ship at all:
| Gate | Evidence required | Launch effect |
|---|---|---|
| U.S. adviser perimeter | Written federal and state counsel opinion on exact paid feature set and marketing | Blocks paid GA and personalized decision features |
| Australia AFSL perimeter | Written Australian counsel opinion on provider entity, overseas users, and exact features | Blocks paid launch and Australian access |
| International markets | Market-specific UK/EU/Canada/Singapore opinions and implementations | Blocks those markets |
| Trademark | Full AU/U.S. clearance or replacement-name clearance | Blocks public brand expansion and material spend |
| Data rights | Executed source-specific rights matrix and fail-closed enforcement | Blocks affected content/features |
| Privacy | Final privacy notice, lifecycle, contracts, DPIA where applicable, security program, DSAR/deletion/export | Blocks real portfolio data and affected jurisdictions |
| Product controls | Automated and manual evidence that forbidden outputs, defaults, labels, colors, and routes are blocked | Blocks activation |
| Marketing | Claims file, disclosure review, beta/testimonial releases, financial-promotion review where applicable | Blocks external publication |
| Independent QA | Staging acceptance against legal acceptance criteria | Blocks production approval |
The opinions must bind to screenshots, prompts, refusal behavior, defaults, terms, data flows, notification copy, marketing copy, and the exact release. A generic memo about "educational software" is insufficient.
Build:
Prepare, but keep feature-flagged off:
Create a legal dossier for each feature: exact UX, prompts, sample outputs, inputs, personalization variables, defaults, colors, notifications, logs, refusal tests, and marketing description.
Activate only after the applicable written opinion, privacy readiness, data rights, independent QA, and human production approval. International markets should be separate feature and policy configurations, not a single global switch.
Answer: Not cleared; direct public-expansion hold. The existing MOAT field in financial services and investment research creates material risk. Obtain counsel clearance or change the name.
Answer: The safe core may fit an impersonal research/publishing posture. The paid personalized concept does not currently have a reliable publisher-exclusion position. Personalized chat, allocation context, alerts, and normative analysis require written federal/state counsel advice before activation.
Answer: Personalize retrieval and organization, not a security conclusion. Portfolio, goals, horizon, risk/loss tolerance, financial circumstances, and behavior must not influence an opinion, priority, warning, score, target, assumption, or action.
Answer: Do not use "broken." Use neutral source-bounded status, objective published criteria, human review, exact context, dates, correction rights, recheck logic, and no intent/credibility inference.
Answer: Use the text in Sections 10 and 11. Disclosures must be contextual and conspicuous. They do not expand what the product may do.
Answer: UK, EU/EEA, Canada, Australia, and Singapore are RED for the current consumer product. Geo-block until local counsel and market-specific product/privacy/marketing controls are complete.
Answer: Free/private is not exempt. Limit beta to invited adult U.S. testers, sample data, no personalized advice, no held content, full terms/privacy, separate testimonial permission, and no public access.
ClassicInvestor may proceed as a source-grounded workspace that helps a user conduct and record their own research. It should not present itself or behave as a system that determines investment quality, suitability, timing, valuation, portfolio fit, or action.
The immediate product path is:
This document should be updated whenever the provider entity, jurisdiction, pricing model, feature behavior, content source, AI provider, portfolio integration, marketing promise, or product name changes.