Summary: ESMA’s Supervisory Briefing on Algorithmic Trading (February 2026)
Introduction
Following common supervisory action (CSA) taken by ESMA (The European Single Markets Authority) in the wake of the Nordic Flash Crash of 2022, it has expressed concern that there is a fragmented understanding of and approach to algorithmic trading across National Competent Authorities (NCAs). This guidance aims to bring convergence in the understanding of algorithmic trading and the supervision of it, across NCAs.
Although non-binding, ESMA expects firms to incorporate the guidance, not least as a tool to assess their own compliance with the regulatory framework surrounding algorithmic trading.
Background: the Nordic Flash Crash and subsequent Common Supervisory Action (CSA)
As a quick reminder of the events that led to that crash: a trade was to be made in an index, with a notional value of US$58 million. However, and apparently inadvertently, the figure of 58 million was inputted into the quantity field of the trading management system. The result was a trade of a basket of equities with a notional value of US$444bn. Although internal controls prevented $255 billion of this ‘fat fingers trade’ from going through, $189 billion in equities were sent to the trading algorithm to be sold in portions throughout the day. In all, about $1.4 billion in equities were sold. These trades in turn triggered a mass sell-off, that saw the OMX Stockholm 30 Index fall by almost 8% in a matter of minutes. At one point, €300 billion in value had been wiped off exchanges.
Algorithims, algorithmic trading strategies and algorithmic trading
The guidance starts - perhaps surprisingly given that algorithmic trading began in Europe in the early 1990s and that ESMA recognises that trading without the use of computer algorithms is the exception nowadays - by taking some time to define and explain two basic concepts namely:
an algorithim as “..a computerised set of instructions or rules that autonomously determines one of more parameters of a trading order…”; and
an algorithmic trading strategy as “…a set of decision logic implemented through one of more algorithms, that autonomously pursues a defined trading objective…”.
As algorithmic trading is defined in Article 4(1)(39) of MiFIDII, ESMA’s guidance looks to supplement that definition with some conceptual explanations, clarifications and examples. ESMA, of course, leans heavily on the core feature of algorithmic trading being “...the automatic algorithmic determination of any individual trading parameter of an order…” and gives typical examples as: the ordering decision; price; timing; quantity; or post order management. However, ESMA goes further and identifies aspects of other activities, such as order generation logic, execution strategy selection and market condition analysis that, although falling outside of basic order processing, will likely qualify as algorithmic trading, if they are automated.
It is fundamentally important that firms appreciate that they will be considered by ESMA to engage in algorithmic trading where the threshold requirement of algorithmic trading is met, notwithstanding that there may be human intervention in, or control of, any part of the relevant trading process.
The obligation to test
Firms must prove their trading systems can handle twice the highest volume of trading reached during the previous six months and must retest after any "material change" to algorithms. Importantly, firms must watch out for minor, creeping recalibrations (especially with machine learning tools) that can accumulate into a material change over time.
Guidance on PTCs
Although this guidance targets algorithmic trading and the concomitant requirement for PTCs, firms are reminded that they are expected to carry out risk assessments and implement and periodically review controls on manual (or non-algorithmic) trading so as to prevent erroneous orders being made. As for PTCs, highlights of the guidance include:
Firms should not rely on PTCs established by trading venues.
PTCs should apply PTcs to all orders regardless of, for example, size and frequency of trading.
PTCs should be established and reviewed on a collaborative basis and involve risk management, compliance, trading and IT functions.
Firms must monitor their algorithmic trading in real-time and this monitoring must include alerts from PTCs.
In relation to real-time monitoring, firms are required to have two lines of defence both at the level of the trader and at the risk management level.
There must be clear procedures in place to be followed once PTCs are reported.
If a firm uses third-party algorithms or outsources its risk management, the firm remains fully accountable for compliance and must maintain operational control to suspend or terminate trading if needed
Firms offering Direct Electronic Access (DEA) to clients bear the responsibility for those clients' compliance. DEA providers must govern and apply PTCs, including real-time monitoring, to all orders submitted by their clients, independently of any controls that clients may have set up themselves.
Firms must implement "hard blocks"—strict default limits on price, value, and volume that cannot be overridden by a trader. ESMA also recommends the use of “soft blocks”, which serve as operational warning alerts that a trader must actively review and override before an order is submitted.
Artificial Intelligence
ESMA notes that while AI-based algorithmic trading is currently excluded from being a "high-risk" use case under the AI Act, it could still be classified as "limited risk" if it interacts directly with natural persons, triggering transparency requirements. Furthermore, compliance staff must be able to understand and adequately explain how AI impacts the algorithmic decision-making processes.
Links:
https://www.esma.europa.eu/press-news/esma-news/esma-issues-supervisory-briefing-algorithmic-trading
https://www.esma.europa.eu/sites/default/files/2026-02/ESMA74-1505669079-10311_Supervisory_Briefing_on_Algorithmic_Trading_in_the_EU.pdf