💡 This Week’s Focus
From LLM models incorporating sentiment analysis into dynamic hedging to yet more on rules, agents and the environment – this week’s roundup dives into what’s new, what’s coming, and what regulators are starting to say louder 📢.
- 🧠 Smart Tech, Smarter Trades
A new study shows how 💬 Large Language Models (LLMs) are being used for real-time sentiment-driven hedging. By scanning news 📰, social media 🐦, and market reports 📊, these models let trading desks adjust positions dynamically in response to market mood swings. The result? 💪 More adaptive, risk-aware strategies with better returns and stronger downside protection.
Read more here – https://arxiv.org/html/2404.09462v1?utm_source=chatgpt.com
Meanwhile, Agentic AI — think 🤯 autonomous systems that not only learn but act without humans — is inching closer to reality in trading, compliance, and ops.
🏛️ More Regulatory Responses
- 📘 ICMA to IOSCO: Clarification – Is AI used in Algos?
ICMA welcomed IOSCO’s work on AI but flagged a key point – benefits and risks of AI in financial services are highly dependent on the type of AI technology used, how it is developed and for what purpose. The main query centered on concerns ICMA has on claims that GenAI and LLMs are used in algorithmic trading—which ICMA members have not yet observed. This could be that ICMA members are financial services firms aware of implications of financial services regulation – not everyone using AI in financial services is under the same level of regulatory scrutiny and this may be real issue moving forward.
ICMA other requests for IOSCO:
- 🧑💼 Keep a “human in the loop” for critical decisions
- 🗂️ Define AI terms globally – current language is inconsistent
- 🔋 Highlight environmental costs of LLMs
- 🧩 Coordinate regulation across borders
Read more here – 🔗 ICMA Response to IOSCO (PDF)
- 🧭 EFAMA’s Take
EFAMA’s members are most worried about:
- 🧍♂️ Human-AI interaction risks (over-reliance, automation bias)
- 🦹♂️ Malicious use (phishing, deepfakes, fraud)
EFAMA’s recommendations are:
- 🎓 Better staff training & internal safeguards
- 🎯 Focus on Autonomous AI Agents as a standalone risk class
- 📚 More investor education to reduce scams
- 🌐 Global standards aligned with MiFID II, DORA & GDPR
Read more here – 🔗 EFAMA Response to IOSCO (PDF)
- 📅 Coming Soon: ESMA’s AI Rules
📢 Alongside the latest European Securities and Markets Authority (ESMA) newsletter focused on the risk of using AI in investing 9 🔗 ESMA Newsletter – February & March 2025 (PDF)) ESMA are getting set to release AI guidance for trading systems by May 2, 2025. These rules will define how AI in algo trading must comply with the EU AI Act. More clarity, more compliance — and likely, more questions 🧾.
- 🌍 The Rising ESG-AI Tension
ICMA also highlighted their concern that training LLMs eats serious power 💡 — and that’s tough to reconcile with ESG targets 🌱. Request from ICMA on further exploration of the environmental footprint of advanced AI systems, is in line with earlier articles such as this one from IFRI published in February looking at the anticipated surge in energy demand from AI and data centers, the subsequent potential strain on power grids and the implications for Europe’s energy transition goals- https://www.ifri.org/sites/default/files/2025-02/ifri_buffard-rochegonde_ai_data_centers_energy_2025_0.pdf
As always, thanks for Reading! Let me know what you liked, didn’t and what you would like to see more of.
Happy Easter for those that celebrate it – enjoy your Sunday!
Rebecca
