AI developing faster than market regulator can make rules, FCA warns

Ever-faster artificial intelligence trading 'bots' may make it harder for regulators to monitor markets and prove when rules are being breached, the boss of the City regulator has warned.
Nikhil Rathi, chief executive of the Financial Conduct Authority (FCA) since 2020, said the rising use of AI in the financial world meant that 'clean markets' could be more difficult to achieve in the future.
'What will clean markets mean in the future with more autonomous agents operating, trading at phenomenal speeds across the globe, and how can you prove abuse in that environment? I think that's something that's going to hit us in the next few years,' he said.
A 'clean market' is one in which prices are set by genuine supply and demand forces, not by cheating or manipulation. In a clean market every participant — whether that is big banks, hedge funds or retail investors — play by the same rules, have timely access to accurate information and cannot secretly distort prices through tricks such as insider trading, spoofing orders or spreading false rumours.

Try Our AI Features

Explore what Daily8 AI can do for you:

Learn with AIFact CheckingText to SpeechAI Summary

Related Articles

Reuters

16 minutes ago

• Reuters

World markets on oil watch as Middle East tensions flare

LONDON, June 20(Reuters) - Brent crude oil is up around 20% so far in June, and set for its biggest monthly jump since 2020 as Israel/Iran tensions flare-up. Although relatively contained, the rise has not gone unnoticed just three years after Russia's invasion of Ukraine triggered a surge in energy prices that ramped up global inflation and sparked aggressive interest rate hikes. Here's a look at what rising oil means for world markets. Oil prices have crept rather than surged higher with investors taking comfort from no noticeable interruption to oil flows. Still, pay attention. The premium of first-month Brent crude futures contract to that for delivery six months later this week rose to a six-month high as investors priced in an increased chance of disruptions to Middle East supply . It remained elevated on Friday. Trading at around $77 a barrel , oil is below 2022's $139 high, but is nearing pain points. "If oil goes into the $80-100 range and stays there, that jeopardizes the global economy," said ABN AMRO Solutions CIO Christophe Boucher. "We are just below that threshold." Traders have an eye on shipping, often seen as a key energy bellwether. About a fifth of the world's total oil consumption passes through the Hormuz Strait between Oman and Iran. Disruption here could push oil above $100, analysts say. Blocked shipping routes would compound any supply shock. Though the big oil producing countries that make up OPEC+ have promised an extra 1.2 million barrels a day, none has yet been shipped or delivered, said hedge fund Svelland Capital director, Nadia Martin Wiggen. Blocked shipping routes would mean this expected supply would not come into the international market, she said. She's watching freight rates closely. "So far, freight rates show that China, with the world's biggest spare refining capability, hasn't started panic buying oil on supply concerns," said Wiggen. "Once China starts to buy, freight rates will rise, and world's energy prices will follow." Rising oil prices raise worries because they can lift near-term inflation and hurt economic growth by squeezing consumption. High oil prices work like a tax, say economists, especially for net energy importers such as Japan and Europe as oil is hard to substitute in the short term. Lombard Odier's chief economist Samy Chaar said that sustained oil prices above $100 a barrel would shave 1% off global economic growth and boost inflation by 1%. Unease rose after Israel launched its strike on Iran a week ago. An initial rally in safe-haven bonds soon evaporated as focus turned to the inflationary impact of higher oil. The euro zone five-year, five-year forward, a closely-watched gauge of market inflation expectations, climbed to its highest level in almost a month . "In the United States $75 oil is enough to, if it's sustained, boost our CPI forecast by about half a percent by the year end, to go from 3 to 3.5%," said RBC chief economist Frances Donald. Turkey, India, Pakistan, Morocco and much of eastern Europe where oil is heavily imported are set to be hit hardest by the rise in crude prices. Those that supply it; Gulf countries, Nigeria, Angola, Venezuela and to some degree Brazil, Colombia and Mexico should get a boost to their coffers, analysts say. A shift is taking place in the dollar. In recent years the currency has risen when oil rallies, but it has had only limited support from oil's latest rise, with a weekly gain of just 0.4% . Analysts expect the dollar's downward trend to resume, given expectations of limited Middle East risks for now and underlying bearish sentiment. It has weakened around 9% so far this year against other major currencies, hurt by economic uncertainty and concern about the reliability of U.S. President Donald Trump's administration as a trading and diplomatic partner. No doubt, a weaker dollar heals the sting from higher oil, which is priced in dollars. "For oil-importing countries, the dollar's fall offers some relief, easing the impact of soaring oil prices and mitigating wider economic strain," UniCredit said. In the absence of an oil-supply shock, world stocks are happy to stick near all-time highs. "Investors want to look past this until there's a reason to believe this will be a much larger regional conflict," said Osman Ali, Goldman Sach's Asset Management's global co-head of Quantitative Investment Strategies. Gulf markets sold off on the initial news, then stabilised somewhat, helped by the higher oil prices. U.S. and European energy shares, particularly oil and gas companies have outperformed (.SPNY), opens new tab, (.SXEP), opens new tab, as have defence stocks. (.SXPARO), opens new tab Israeli stocks, (.TA125), opens new tab up 6% in a week, have been the most notable outperformer. Stocks of oil consumers have been the worst hit, airlines stand out.

Sky News

16 minutes ago

• Sky News

'Inflation and customer cutbacks' blamed for big dive in retail sales

Retail sales volumes suffered their largest monthly fall since December 2023 last month, according to official figures which suggest a link to rising bills. The Office for National Statistics (ONS) reported a 2.7% decline in the quantity of goods bought in May compared to the previous month. The body said its interaction with retailers suggested " inflation and customer cutbacks" accounted for the fall, which was across all categories, but led by food. The seasonally adjusted data - which reflects the effects of holidays - means that those for Easter are modified to give a clearer picture of sales trends. A poll of economists by the Reuters news agency had expected to see a decline in volumes of just 0.5% in May following April's growth of 1.3%. May was the month when households would have noticed the hit from the so-called 'awful April' above-inflation hikes to essential bills, including council tax, water, mobiles, broadband and energy. Retail sales growth had proved to be resilient this year until May but April brought a number of additional curveballs to confuse sentiment and place pressure on the economy generally. 2:01 Donald Trump's "liberation day" tariff regime kicked in while Budget measures, including rises to minimum pay levels and employer national insurance contributions (NICs), also placed additional costs on businesses. Retail is the UK's largest private sector employer. It had threatened higher prices and hits to hiring and wage growth ahead of the tax take coming into effect. While the inflation picture for May was largely flat, the ONS reported last week employment data showing a tick up in the unemployment rate to 4.6% in the three months to April. Figures from the taxman also showed a 109,000 decline in payrolled employment during May. Further data from the ONS on Friday revealed a £1.8bn jump in additional "compulsory social contributions" - largely made up of NICs - in May. It took the extra tax take to a record £30.2bn across April and May but borrowing still surged to £17.7bn last month, the second highest figure on record for May, the ONS said, as the chancellor bids to grow the economy within tight fiscal rules. Consumer spending accounts for around 60% of UK output. A closely-watched measure of consumer confidence covering June showed no rise in consumers' expectations for spending on so-called big ticket items. The GfK survey was taken after the UK's trade truce with the US but before Israel's air war with Iran began. That has pushed oil and natural gas prices up by double-digit percentage levels in under a week, threatening a new energy-led cost of living threat. It's another challenge that retailers, businesses more widely, and Rachel Reeves could do without. Thomas Pugh, economist at audit firm RSM UK, said: "Looking ahead to the budget in the autumn, the under performance of the economy and higher borrowing costs mean the chancellor may already have lost the £9.9bn of fiscal headroom that she clawed back in March. "Throw in the tough outlook for many government departments announced in the spending review and U-turns on welfare spending and the chancellor will probably have to announce some top-up tax increases after the summer."

Finextra

24 minutes ago

• Finextra

Inside AI Assisted Software Development and why tools are not enough (Part 1): By John Adam

The recent squeeze on funding and margins is by no means only being felt in the financial services and fintech sectors. But it's fair to say the pinch is particularly hard and the necessity to quickly and effectively innovate is simultaneously more pressing than ever. The good news is, new AI tools can speed up delivery and improve the quality of software projects without adding to headcount. But even if that general statement is true, just using tools is not enough. Especially in a regulated industry like financial services. If there is no pre-approved list of tools and how and where they are applied in an SDLC (software development lifecycle), organisations have governance, observability, measurability and consistency issues. If 'real' gains are not measured by benchmarking against 'before', do they really exist? Tree falling in a forest metaphor. Certainly not in a way that can be scaled across or up an organisation. There is no clear business case, just intuition. Are tools and where and how they are being used compliant with organisational policy and regulatory frameworks? Has anyone read the privacy policies? I'm personally convinced that a big AI company having its Facebook/Cambridge Analytica moment falls under 'when, not if'. And when the first big AI privacy scandal does break, you don't want your organisation published in a list in a newspaper. To benefit from and scale the gains of an AI-assisted SDLC, organisations need a framework for structured, consistent integration + governance, observability and measurability. Just tools isn't enough. Realistic gains from an AI-assisted SDLC It's important to note that at the time of writing, we are in a period of rapid change in AI tooling. A good framework operates at a level or two higher than specific tools and allows for them to be interchangeable with upgrades. The market most of us operate in is at a point in its cycle where resources are at a premium. Most of the organisations I work with are expected to deliver more with less compared with pre-2023. In that context, banking the productivity gains achievable with AI tooling is non-negotiable. Organisations are demanding it in the demand for greater, better output despite fewer resources. Getting it right is also non-negotiable and that means marrying increased productivity with measurability, observability and governance, which I cover in-depth in Part 2 of this article. As an introduction to building a proper framework, I'll start by explaining the realistic improvements AI can provide to each stage of the SDLC: Product prototyping Developers use prototypes to test idea viability and functionality, and to gather user and investor feedback. Historically, the average prototype required 2 to 6 weeks of teamwork to complete. But by amplifying developers' work via low-code/no-code prototyping and AI-generated code and other AI tools, a clickable prototype can now be completed in days or even hours. UX/UI design UX (user experience) and UI (user interface) designers collaborate closely with developers to design website and app interfaces. Using AI tools that can quickly generate multiple design mock-ups and UI components based on foundational style guides and example concepts, designers can visualise ideas and user flows in various contexts to improve design clarity and direction long before designs touch a developer's desktop. Clarity improves the quality of initial designs and reduces designer-developer back-and-forth, meaning larger projects that took 4 to 6 months to complete now require far less effort and time. Even UXR (User Experience Research) is accelerated and refined. User interviews are, by necessity, long and complex, and result in large, qualitative datasets. AI tools can highlight patterns and repetition in datasets and transcripts in seconds—shining a spotlight on insights, false positives or even biased questions that human researchers may have overlooked. Architecture Software architects plan higher-level design, bridging technical and business requirements. Their diagrams include the sum of a products' components and their respective interactions; until recently, the initial design phase alone took 1 to 2 weeks. Using AI, architects can quickly draw up diagrams to easily visualise these relationships and standardise dependency versions across services. AI can also be trained to use PR comments to report architectural violations, and libraries can be unified to encourage stability across features. Better consistency and immediate feedback mean architects can work faster and create fewer iterations of a product before diagrams meet stakeholder expectations. Coding AI-powered tools for coding have a variety of use cases. My team uses a mix of tools and GenAI to: ensure comprehensive project documentation, automate code documentation and README generation, scan for duplicate code and suggest improvements, improve understanding of complex, inconsistent or unfamiliar code bases, unify code styles and standards across different microservices, and perform code completion and check for bugs and inconsistencies based on defined standards. Paired with manual oversight to catch any mistakes, we've accelerated writing and testing code by a minimum of 20% across projects. GenAI makes complex codebases easily understandable—meaning team members can flexibly move to work on unfamiliar projects and diminish time spent on internal comms by about 25%. One tool we use is SonarQube, which reviews code without executing it. It runs automatically in GitLab CI/CD (Continuous Integration/Continuous Delivery and Deployment) pipeline to find bugs, report security vulnerabilities, and enforce code standards to unify style and mitigate potential misunderstandings down the line with better code readability. Testing and QA (Quality Assurance) As they write code, developers write and run unit tests to detect initial bugs and security issues that eat up between 10% and 20% of their time. The SDLC is slowed further by code reviews and PRs, or feedback from experienced colleagues. Tests are postponed by days, sometimes weeks, if various code reviews are required and dependent on busy colleagues. GenAI can augment developers' efforts by writing unit tests, conducting code reviews and PRs in real time, and automatically generating and solving for edge cases to overcome bottlenecks like a lack of expertise or teammates' availability. AI augmented QA can reduce redundancy, unify access to code, and consolidate fragmented knowledge across a project to make a QA team more efficient. And AI-driven tools like Selenium, for example, can automate web app test writing and execution, accelerating product releases and improving product reliability. Automated testing is especially compelling in the context of projects with tight deadlines and few resources. For example, my team's AI toolkit for QA testing includes Llama 3.3 LLM to generate test cases and analyse code and Excel-based legacy documents, IntelliJ AI Assistant to automatically standardise test case formatting, and GitLab to run and test scripts automatically in the CI/CD pipeline. QA is one of the most impactful applications of AI tools in the SDLC and can commonly slash the resources required by up to 60%, while increasing test coverage. Deployment When a product is deployed to end users, AI can be added to the CI/CD to forecast use patterns and improve caching strategies, as well as automatically prioritise and schedule tasks for parallel execution. With AI oversight, the number of repetitive tasks is automatically reduced and resource allocation anticipated, improving latency and product release cycles without added manual effort. And AI-driven caching accelerates and simplifies rollbacks (reverting a newly deployed system to a more stable version of itself) by analysing previous deployments and predicting the necessary steps, reducing further manual effort by DevOps teams, for instance. My team uses Dytrance during deployment, which monitors and analyses system status, and sends self-healing recommendations in real time. Maintenance and Monitoring At this stage, teams work to fix bugs, keep the system secure and functioning well, and make improvements based on user feedback, performance data and unmet user needs. AI can automatically perform root cause analysis for error monitoring, and suggest solutions for maintenance and debugging. Tools my team uses include AWS Cloud Watch and Azure Monitor with AIOps, which automatically collect, analyse, and suggest responses based on monitoring data, accelerating issue response and system updates by 10x. The big picture The acceleration of the individual stages of software development is incentive enough for some teams to add tools and GenAI models to their workflows; especially at stages like QA and coding, where use cases are various and results potent. But by taking a step back and considering AI's impacts on the SDLC holistically, the argument in favour of AI implementation can be turned into a real business case. A business case that can be used to accelerate AI transformation across an organisation: Backed by a strong framework, organisations implementing AI across their SDLC see a 30%+ acceleration across projects in the first 6 months. The keyword being 'strong.' Organisations need a framework that guides leadership to select tools and govern their use, measures outcomes to understand the amount of value different tools offer, and encourages adoption in teams' workflows. Without it, teams are unable to measurably extract the full potential from new tools and efforts, and risk breaching internal and third-party governance in areas such as data privacy. Keeping my word count and your patience in mind, I split my deep dive into a framework for AI governance, measurement and adoption into a separate article: Here is Inside an AI-assisted software development framework: using tools is not enough Part 2.