The AI Act starts biting, and the trader still owns the plan
On 2 August the European Commission began enforcing the AI Act's high-risk obligations, and UK supervisors sharpened their language on stress testing AI systems. Olga Magomedova on what the tightening rules do, what they do not do, and where the responsibility still sits.
A supervisory line was drawn this month
The rulebook moved from theory to enforcement. From 2 August 2026 the European Commission's AI Office, together with national authorities, began enforcing the AI Act, and new transparency rules started to apply, requiring certain AI systems to tell users when they are interacting with AI and when content has been generated or altered by it. That is a supervisory event, not a marketing one. It changes what firms must document and what users are entitled to know.
In parallel, UK supervisors moved in the same direction with different vocabulary. The FCA is examining whether existing frameworks, including the Consumer Duty, the Senior Managers and Certification Regime, operational resilience requirements and the Critical Third Parties regime, remain suitable for an AI-enabled financial sector, and its review warns that autonomous and interconnected AI may amplify existing risks and that supervising AI-enabled firms will require deeper technical expertise and new forms of testing, assurance and continuous monitoring.
What the new rules actually change for the individual trader
For a firm deploying models, the change is operational. Logs, monitoring, and documentation are now duties, not aspirations. For the individual trader using a retail platform, the change is smaller in surface area and larger in principle. The tool now has to tell you it is a tool.
Chatbots and other interactive AI systems will have to tell users they are dealing with AI rather than a human, deepfakes will have to be labelled, and AI generated or altered content will have to carry machine readable marks so it can be detected more easily. That is a floor, not a ceiling. Disclosure lets the trader see the tool. It does not tell the trader what to do with what the tool says.
The Bailey line and why traders should read it
One line from the Bank of England governor is worth attention. "As a regulator we require that banks prove to us that they can do this through stress tests and penetration testing," Bailey wrote. The sentence was written for banks. The habit belongs to any trader operating an AI assisted process.
The engineering translation is direct. A model that has not been stress tested is a model whose failure modes are unknown. A trading plan that relies on such a model is a plan with an unmeasured tolerance. For Magomedova, the correct posture is the one the regulator is asking of institutions. Prove, with artefacts, that the system does what it is claimed to do under conditions it was not built for.
Regulation supplies scaffolding, not judgement
The temptation, when supervision tightens, is to assume the supervisor now carries the risk. The supervisor does not. The trader still holds the position. Regulation raises the floor beneath which firms cannot go. It does not raise the ceiling of any individual trader's discipline. That work remains private.
The AI Act's transparency duties help the trader in one specific way. They make the tool visible as a tool. The rest of the work, sizing the position, holding the plan, writing the post-mortem when the model was wrong, sits on the same desk it always has.
AI is a powerful tool. But tools do not replace responsibility.
The habit the rulebook implicitly recommends
Read the AI Act obligations as an inventory. Logs of decisions. Post-market monitoring. Records of when the model behaved unexpectedly. That inventory looks familiar. It is the same inventory a disciplined trader keeps, at smaller scale, in a personal journal.
The trader who kept that record for years before any regulator asked for it will find the new rules unremarkable. The trader who did not will discover that supervision has produced, in its clumsy institutional way, the habit that would have paid on its own.
The principle that does not move
For Magomedova, the news of the month does not change the underlying rule. The rule is that every position must be one the trader can defend in their own words, whether or not a model contributed to it, and whether or not a regulator has now required the model to identify itself. The disclosure is welcome. The responsibility is unchanged.
The AI Act will produce more documentation across the industry. Some of it will be useful. Some of it will be theatre. The private ledger, kept honestly, remains the artefact that decides whether a trader survives the next regime the model has not seen.