Market-structure

Indonesian multi-asset platform Pluang launched agentic trading in controlled beta, letting whitelisted users connect ChatGPT, Claude or Gemini directly to live regulated accounts to research, propose and execute trades across stocks, gold, money market instruments and crypto, with public rollout set for mid-August

Every AI-proposed transaction still requires explicit user confirmation before execution, and cofounder Claudia Kolonas said safeguards "must be embedded at infrastructure level, not depend only on model algorithms." The move follows similar 2026 launches from Robinhood, Interactive Brokers and TradeStation in the US, and puts Indonesia among the first Southeast Asian markets to open a regulated brokerage's execution layer to third-party general-purpose AI assistants.

· Source: The Jakarta Post


The structurally interesting part of this launch isn't that Pluang added an AI feature, brokers have been bolting AI research tools onto trading apps for two years now. It's that the assistant sits on the client side of the API and the user brings their own model, ChatGPT, Claude or Gemini, rather than Pluang building and owning a proprietary trading assistant. That's a different architecture from Robinhood's or Schwab's approach, where the AI layer is the broker's own product. Pluang is instead exposing an execution API that any general-purpose assistant can call, with the confirmation gate as the safety boundary.

That architecture choice has real market-structure implications once it scales past a whitelist. A brokerage account reachable by any user-supplied LLM is a fundamentally different attack surface and compliance problem than one reachable only through a vendor-controlled assistant the broker has tested and can update unilaterally. Kolonas's line about safeguards needing to sit "at infrastructure level, not depend only on model algorithms" reads as a direct acknowledgment of that: if the safety net depends on which model a user happens to be running, the broker has no control over the weakest link.

The confirmation-gate design, AI proposes, human approves, execution triggers only after that approval, is the same shape regulators in the UK and US have been pushing toward all year under various human-in-the-loop frameworks. What's notable is that Pluang is implementing it as a default architectural constraint from launch, in a jurisdiction, Indonesia, that doesn't yet have an AI-specific trading rulebook comparable to what FINRA or the FCA have been drafting. That's a case of market practice arriving ahead of formal regulation rather than in response to it.

Watch the mid-August public rollout for two things: how Pluang handles the first mass-market failure mode, a user's AI assistant proposing a trade the user didn't intend to approve, and whether Indonesia's OJK follows with formal guidance once the feature is live rather than beta-gated. A regulated brokerage opening its execution layer to third-party AI models at consumer scale, before the local regulator has written specific rules for it, is exactly the kind of gap that either becomes the template other emerging-market brokers copy or the case study regulators point to when they eventually do legislate.


Read the original: The Jakarta Post - Indonesian multi-asset platform Pluang launched agentic trading in controlled beta, letting whitelisted users connect ChatGPT, Claude or Gemini directly to live regulated accounts to research, propose and execute trades across stocks, gold, money market instruments and crypto, with public rollout set for mid-August. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.