Autonomous models are taking on more active roles across platforms; Binance is allowing agents to trade crypto, Slack is integrating coding agents as channel members, and major banks are deploying Ant International's currency forecasting model.
Binance launched Agent OS, allowing AI tools such as ChatGPT, Claude Code and Cursor to execute trades on the exchange through permissioned sub-accounts. Withdrawals are blocked by default, and daily limits are capped at 50,000 dollars for regular swaps, 100,000 dollars for DeFi transactions and 20 dollars for payments. Users can set whether each order requires manual approval or runs autonomously, while the exchange connects the system via the Model Context Protocol.[1]
Slack introduced Slack Code, rolling out project-specific channels where a human team can tag AI coding agents to collaborate, and the agent opens a dedicated task channel. The channels compare code changes, preview HTML output before shipping, and automatically archive themselves after the task concludes while keeping a complete audit log. Supported partners include Claude Code, Devin, Vercel Agent and GitHub Copilot, bringing automated programming workflows directly into the primary communication platform.[2]
Ant International announced that Barclays, Citi, Deutsche Bank and Standard Chartered are running version 2.0 of its Falcon Time-Series Transformer model for cross-border payments. The model forecasts cash flow and foreign-exchange exposure. The company claims the system achieves a mean absolute scaled error metric at 0.666 and maintains its forecast accuracy above 93 per cent. Barclays runs the model inside BARX NetFX, and Deutsche Bank uses it for liquidity management.[3]
