Capital Markets & Wealth

For broker-dealers, investment banks, market makers, asset managers, wealth managers, robo-advisers and trading platforms, and the vendors whose algorithms they run, and for the clearing brokers, boards, investors and counsel who rely on their decisions: an independent finding on one named AI decision, against the rule that binds it.

The decisions we examine

  • Orders – what an algorithm sends to a trading venue, at what size and price, and whether it can be cancelled at once.
  • Pre-trade controls – which orders are blocked before they reach the market.
  • Surveillance – which alerts for possible market abuse are closed or escalated.
  • Advice and suitability – which products are recommended to a client, and whether they suit that client.
  • Portfolio decisions – where AI signals shape what a fund or model portfolio buys and sells.
  • Onboarding – which clients are accepted, flagged or rejected.

What is changing

  • European Union: an investment firm that trades algorithmically must be able to cancel immediately, as an emergency measure, any or all of its unexecuted orders on any or all trading venues, and must be able to identify which algorithm and which trader, desk or client is responsible for each order (Commission Delegated Regulation (EU) 2017/589, Article 12; applies since 3 January 2018). The United Kingdom keeps the same requirement in its own version of the regulation, alongside the FCA’s rules on algorithmic trading (FCA Handbook, MAR 7A).
  • European Union: ESMA’s statement on AI in retail investment services says firms’ decisions remain the responsibility of their management bodies, that the duty to act in clients’ best interest applies whatever tools a firm adopts, and that firms should keep comprehensive records of their use of AI (30 May 2024; ESMA statement).
  • United States: a broker-dealer with market access must have risk controls reasonably designed to stop erroneous orders, and orders beyond pre-set credit or capital limits, before they reach the market (SEC Rule 15c3-5).
  • Hong Kong: the SFC treats generative AI used for investment recommendations, advice or research as high-risk, and expects a person to review its output before it reaches the client (circular of 12 November 2024; legal summary).
  • Singapore: the MAS Guidelines on AI risk management take effect on 7 October 2027 for financial institutions, capital-markets firms included (Calendar).
  • India: SEBI makes the broker the principal for every algorithm offered through its systems, requires the exchange’s approval for each algorithm, and keeps the exchanges’ kill switch for orders from a given algorithm; the framework applies to all stock brokers from 1 April 2026 (SEBI circular, 4 February 2025; timeline, 30 September 2025).
  • The Calendar also lists guidance that reaches investment and wealth firms in Malta, Mauritius, South Korea, Switzerland, Taiwan, Bermuda and South Africa, each with its source.

These are short summaries of each rule, each linked to its source; they are not legal advice. A finding names, at intake, the exact rule it measures against.

Who asks for the evidence, and who pays

  • The firm that runs the algorithm or the advice tool commissions and pays for a finding on its own system; the chief risk officer or head of trading risk usually signs.
  • A clearing broker or prime broker, which answers for its clients’ orders, may commission a finding on a client firm’s algorithm, with that firm’s written agreement. The reviewed firm counts as a Fijishi client (Governance).
  • Its board or audit committee may commission the finding itself.
  • An algorithmic trading, surveillance or advice-engine vendor may commission a certificate on its own product, which every client can check by number.
  • Investors and acquirers in brokers, platforms and managers commission a Portfolio Screening or a Deal Review.

What each role finds here

  • Head of electronic or algorithmic trading: a tested answer to whether every order from one algorithm, one desk or the whole firm can be cancelled at once, and how fast – Example certificate 3.
  • Chief risk officer and head of trading risk controls: whether pre-trade controls block or only alert, and whether every override is recorded.
  • Chief compliance officer: whether each AI recommendation met the suitability or best-interest rule, and what an examiner will ask first.
  • Head of surveillance: who answers for an alert the model closed, and whether the decision can be reconstructed.
  • Head of wealth or digital advice: whether the advice tool knows the client well enough for the rule, and whether an adviser can override it.
  • Clearing and prime brokers: evidence about a client’s algorithm, from someone with nothing to gain.

Which product answers which question

The proof you can check

  • Example certificate 3: a Kill-Switch Certificate on an investment firm’s execution algorithm, against the EU’s kill-functionality requirement, with a live test of the stop.
  • FWD-1, Stopping an AI decision: our draft criteria on who can stop an AI decision, open for public comment until 15 January 2027.
  • The Regulatory Deadline Calendar: the capital-markets rows, each linked to its source.
  • Example certificate 2 shows how a Baseline Review reads against a supervisory standard, criterion by criterion; a review of an advice tool against a suitability rule takes the same form.
  • The Fijishi Jurisdiction Index covers insurance. For capital markets and wealth, the proof is the Calendar, the example certificate and the method in public (How certification works).

What existing work covers, and what it does not

Management-system certificates such as ISO/IEC 42001 show that a firm manages AI through a documented system. Model validation and testing show whether a model works. Each has real value. Fijishi answers the question that comes after them: for one named decision, does it meet the rule that binds it, and can the firm show who was accountable and who could stop it? What already exists, and what each covers

Start

Write to protocol@fijishi.com with the AI system, the rule it must meet and any deadline. We reply within 5 business days. Intake has no fee; scope and fee are fixed in writing at intake, before any work begins. Both founders examine and sign every finding (Founders). Fijishi sells no insurance or other financial products, builds or sells no AI, and does no consulting work, so it has nothing to gain from what a finding says. Start an engagement

Insurance · Banking & Lending · Capital Markets & Wealth · Who a Fijishi finding is for · Start an engagement