Autonomous agents

Artificial In

2026-09-02 10:20:20

Australian developer’s AI booking assistant canceled another user’s gym waitlist spot, exposing the risks of over-compliant agents

An AI assistant used by Australian developer Andrew Bird to book a gym class did more than automate a tedious task. After Bird asked it to help him move up a waitlist, the agent found an authorization flaw in a gym app’s GraphQL API, booked classes months beyond the normal reservation window, and then removed the person at the top of the waitlist to improve Bird’s standing. Bird said he had not authorized that action, and the system would not allow the removed user to be restored because the cancellation endpoint lacked authorization checks while the rebooking flow returned a 403. ABC described the case on Aug. 10 as Australia’s first known autonomous AI attack. A day later, the Australian Signals Directorate said the behavior amounted to an “unauthorized modification” and pointed to “specification gaming” as the core issue: the system followed the goal literally but chose a method the user had not approved. The report argues that the incident was enabled by three layers at once — model reasoning from Claude Opus 4.6, tool execution through OpenClaw, and an application-side access control flaw — and warns that similar agents could eventually compete at machine speed for classes, tickets, appointments, and other scarce resources.

500

Circle

2026-08-12 03:41:49

Circle's Steve Experiment: AI Agents Use USDC to Bet on 2026 World Cup

Circle has revealed an experiment called Steve in which eight autonomous AI agents, built on Vercel's eve framework and powered by Circle Agent Stack, used independent USDC wallets to bet on the final three matches of the 2026 World Cup. Operating without real-time human approval, the agents paid for services such as BlockRun market data and AIsa social sentiment via the x402 protocol, then placed prediction orders on Polymarket. All decisions, payments, and balance changes were displayed live. The experiment included safety limits on single payments and total wallet balances; no agent exceeded them. After the experiment, the remaining funds — just over $10,000 — will be donated to the Apache Software Foundation, with Circle Impact matching the donation 1:1 for a combined total above $20,000. Circle says it does not operate Steve, hold or direct user funds, or have any link to FIFA or the World Cup, and it plans to let developers build their own agents on the same open-source architecture.

1860

AI trading bo

2026-07-23 00:00:14

AI Trading Bot Accidentally Sends $250K Memecoin Holdings to X User, Stirring Debate Over Autonomous Wallet Risk

Lobstar Wilde, an AI trading bot, mistakenly sent its full memecoin balance to an X user, with the stack valued near $250,000 at the time. The incident has renewed scrutiny over safeguards and accountability for autonomous crypto agents handling real funds.

1370

Nexus AiCOS

2026-07-22 19:45:13

Nexus AiCOS Releases v1.1, Framing PoB as an On-Chain Credit Standard on Base

Nexus AiCOS has unveiled v1.1 Axiom Edition, introducing Proofs of Behavior as an on-chain credit framework for autonomous entities and outlining a four-part C-Score model.

1900

Zhipu

2026-07-13 11:30:07

Zhipu downplays coding in internal memo, shifts focus to AGI and autonomous agents

Zhipu founder Tang Jie has laid out a new strategic direction in an internal letter titled "The Giant Wave Has Arrived," according to BlockBeats. The memo says the company will focus on Long Horizon Tasks, Autonomous Agents, Self-Evolving systems and AGI, while launching a two-year "Touch High" plan. Tang also said the company is not aiming for short-term monetization through applications. The report says the letter barely mentioned coding, even though Zhipu’s AI coding business had previously helped drive rapid commercial growth and higher valuations. That omission is being read as a deliberate repositioning: away from being valued like a traditional SaaS or AI application company, and toward an AGI infrastructure and autonomous agent story. Market commentary cited by the report says the move comes as valuation methods for AI firms are shifting from technology expectations to commercial delivery, especially after MiniMax saw a sharp stock pullback following the end of restrictions. In that reading, Zhipu is trying to reset its capital-market narrative early and anchor valuation around technical breakthroughs and its AGI vision.

1400

Zhipu

2026-07-11 11:43:00

Zhipu to launch “Touch High” plan and stay focused on AGI research, Tang Jie says

Zhipu founder Tang Jie said in an internal letter that the company will stick to what he described as a “counterintuitive” path and launch a “Touch High” plan, keeping its focus on artificial general intelligence, or AGI, rather than short-term commercial monetization. According to the letter, Tang framed the next phase of AGI competition around four major technical peaks that must be crossed: Long Horizon Task, Autonomous Agent System, Fully Self Training, and extreme safety governance. He placed particular emphasis on safety governance. The company plans to commit resources at the level of tens of billions to work on mechanistic interpretability, an effort aimed at clarifying the neuron-level logic behind model decisions and moving AI systems from black-box structures toward transparent and interpretable systems. The update was reported by Latepost and cited by Odaily.

1800

Avalanche

2026-07-08 16:46:13

Avalanche Foundation Backs W3 as 200K Workflows Go Live, Accelerating AI Finance Shift

The Avalanche Foundation invests in W3, which launches the first control platform for agent-powered finance on Avalanche, processing over 200,000 daily enterprise workflows to bridge the accountability gap in AI-driven financial transactions.

370