All Things AI in Trading – 21st July 2025
Here’s what I’ve learnt this week
How AI is reshaping markets, redefining finance infrastructure, and accelerating regulatory and governance shifts – from Claude’s institutional launch and tokenised collateral breakthroughs to the rise 24/7 trading and EU/UK AI compliance frameworks – here’s what I’ve learnt this week on AI in Trading.
- AI reshaping Markets – The STOXX 600 staged a notable mid-week rally last week, climbing 0.7% on July 17 as robust earnings from major industrial players like ABB helped reverse earlier declines (Reuters).
Although European markets in 2025 generally echo the calm seen before the pandemic, they are now more reactive to specific events… initiatives like LSEG’s exploration of 24/7 trading point to a market increasingly shaped by automation and AI (Financial Times). - Claude takes on BBG and AI moves one step further into finance – Anthropic’s launch of Claude for Financial Services signals a pivotal shift…
Integrated with S&P Global, FactSet, Snowflake, PitchBook, and Databricks, Claude builds models, benchmarks sectors, and delivers structured answers.
Early adopters include Bridgewater’s AIA Labs, NBIM, AIG, and Commonwealth Bank of Australia. In tests, Claude Opus 4 solved 5 of 7 Financial Modelling World Cup challenges with 83% accuracy. - Tokenised Collateral and On-Chain Settlement – Aberdeen, Lloyds Banking Group, and Archax recently completed the UK’s first live use of tokenised collateral in an FX trade.
Meanwhile, Citi is preparing a stablecoin and tokenised deposit platform. With the passing of the Genius Act, regulated banks may soon operate 24/7 on-chain settlement rails. - UK Government’s AI Coding Guidance Sets a Governance Benchmark – The UK Government Digital Service has published structured guidance on AI coding assistants.
Though aimed at government teams, it sets expectations of transparency, validation, and secure deployment that will resonate with financial services. - EU AI Act Moves Closer to Implementation – The Commission published a final draft code of practice for GPAI models.
OpenAI indicated it will sign, while Meta declined, citing regulatory overreach.
The Commission also released guidance for systemic AI models (Google, OpenAI, Meta, Anthropic, Mistral), mandating adversarial testing, incident reporting, and strong cybersecurity.
As always, thank you for reading! Let me know what you found helpful, what you disagreed with — and what you’d like to see more of.
Many thanks,
Rebecca
