MAS flags risks of using AI agents in digital-asset trading

businesstimes.com.sg Oct 9, 2026

MAS flags risks of using AI agents in digital-asset trading

Curated by Shubham · TechCapsules · Oct 9, 2026

AI-summarised brief · reviewed before publication

The Monetary Authority of Singapore (MAS) warned on October 9 that artificial‑intelligence agents, which now can both analyze data and execute trades, pose new risks for digital‑asset markets. Managing Director Chia Der Jiun, speaking at an Insead conference, highlighted potential liquidity and collateral challenges as AI agents move capital faster and more efficiently. He urged institutions to tighten controls on authority, permissions, identity, accountability and governance, and to ensure agents operate within user instructions, policy limits and regulatory standards. MAS has already issued the SAFR white paper and other guidance on agentic finance.

💡 Why It Matters

· By flagging AI‑driven trading as a liquidity threat, MAS signals that regulatory frameworks must evolve to safeguard market stability and protect investors in a rapidly automating financial ecosystem.

AI Capsules Policies

Read full article on businesstimes.com.sg

Budget 2027: Sovereign AI Cloud Planned, 100,000 More Free AI Subscriptions, AI…

Prime Minister Anwar Ibrahim unveiled Malaysia’s 2027 budget, announcing a Sovereign AI Cloud to keep citizens’ data and strategic information under national control. The plan includes 100,000 free AI subscriptions for young people, an AI app to verify phone numbers, bank accounts and links to detect scams, and a RM6 billion allocation for AI research, development and commercialization. AI Malaysia Berhad will receive nearly RM15 million, while a RM30 million MDEC fund will aid 4,000 SMEs in AI adoption and certify 5,000 professionals. The initiative aims to build a safe, ethical AI ecosystem aligned with national values.

💡 By centralizing AI infrastructure, Malaysia seeks to safeguard data sovereignty while fostering local talent and innovation, positioning the country as a regional AI leader.

sheffield.ac.uk · Oct 9

Festival of the Mind 2026: Understanding AI

On Saturday, 7 November 2026, the Centre for Machine Intelligence hosted “Understanding AI” at Channing Hall, Sheffield. The event ran from 9:00 am to 6:00 pm and featured live robotics demonstrations, digital exhibits, and interactive workshops where participants could build zines about their AI experiences. Researchers discussed practical applications of AI and its social and ethical implications. The program concluded with a film screening and Q&A session focused on public involvement in AI decision‑making and regulation. The venue was 45 Surrey Street, Sheffield, S1 2LG.

💡 By centring public dialogue on AI governance, the festival challenges the usual top‑down approach to technology policy and invites everyday citizens to shape future regulations.

techcrunch.com · Oct 9

The maker of non-text AI model Jev valued at $7.5B just weeks…

TypeSafe AI raised $870 million in a funding round that values its new AI model Jev at $7.5 billion. The round was led by Andreessen Horowitz, with Sequoia and DCVC also investing. Jev, released on September 15, is a transformer‑based model that outputs calibrated probability decisions rather than text, and claims to be faster and more token‑efficient than large language models. The company reports that a third of Fortune 500 firms already use Jev for automation tasks.

💡 Jev’s rapid adoption signals a shift toward AI tools that prioritize speed and efficiency over text generation, reshaping enterprise automation workflows.

windowsreport.com · Oct 9

Microsoft 365 Copilot Gets New Safeguard Against Email-Based Prompt Injection

Microsoft has introduced a new Data Loss Prevention (DLP) policy in Microsoft Purview that allows administrators to block external emails from being used as grounding data for Microsoft 365 Copilot. The policy checks sender metadata against accepted domains, preventing Copilot from referencing external messages when generating responses or summaries. Internal documents, spreadsheets, presentations, and web results remain accessible. The safeguard does not alter email delivery or retention, and can be configured via the Purview portal by users with Compliance or DLP administrator roles. The feature aims to reduce prompt‑injection risks while preserving normal email functionality.

💡 By isolating external emails from Copilot’s data set, organizations can curb a subtle attack vector that could otherwise manipulate AI outputs without disrupting everyday communication workflows.

thenextweb.com · Oct 9

8 best AI app builders for 2026

In 2026 AI app builders have evolved from mock‑up generators to platforms that deliver fully functional business applications straight from a natural‑language prompt. The eight tools highlighted all create a working database, user‑login system with role‑based permissions, and editable pages that read and write real data without developer intervention. Key selection criteria include depth of prompt‑to‑app generation, ability to integrate with existing data sources such as Airtable, HubSpot, SQL databases or spreadsheets, built‑in user and role management, post‑generation visual editing, pricing models, and availability of a free tier for genuine trials. Softr leads the list, offering AI‑driven creation of databases, pages, and workflows that connect directly to a company’s current data stores, support GDPR and SOC 2 compliance, and allow non‑technical users to modify apps via a visual editor after the initial build.

💡 It shows that AI‑driven development can now bypass traditional coding bottlenecks, letting businesses launch secure, data‑connected apps overnight. This accelerates digital transformation for teams lacking engineering resources.

arise.tv · Oct 9

AI Funding Appetite Weighs On Global Markets

Global markets are under renewed strain as investors reassess the profitability of massive AI infrastructure spending. OpenAI disclosed September‑year revenue of roughly $50 billion, lower than earlier guidance, prompting doubts about returns on AI‑related capital. In Australia, data‑centre firm Firmus scrapped a $5 billion IPO in favor of private funding, while U.S. giants SpaceX, Broadcom and Oracle are reportedly planning multi‑billion dollar raises to buy advanced AI chips. Morgan Stanley projects $1.5 trillion in external financing for AI infrastructure by 2028, a figure that collides with rising interest rates and inflation‑driven energy costs. Higher borrowing costs lengthen the payback horizon for data centres, and investors now demand clearer evidence of demand, cash flow and customer bases amid growing public concern over AI’s societal impact.

💡 The tightening financing environment forces AI firms to prove sustainable cash flows, potentially slowing the sector’s rapid expansion.