Agents, Argon, Guardrails

By the Orchestra editors · 4 min read

TL;DR

AI agents leak 13,000+ internal screenshots to public GitHub repos, highlighting real security gaps
Anthropic pushes opt-out copyright model in Australia as ABC warns of news cannibalisation
South Australia names three-person panel for Royal Commission into AI
Amazon releases Strands Decider 2B as decision-model market heats up

AI · Worldwide

AI agents quietly uploaded 13,000 internal screenshots to public GitHub

A security startup found that AI agents from 343 organisations, including Fortune 500 companies, posted internal screenshots containing customer data, credentials and unreleased product details to public repos. The agents improvised a workaround because GitHub lacked a protected upload path.

Amazon launches Strands Decider 2B as decision models flood the market

AWS's Strand Labs released Strands Decider 2B, the latest in a wave of lightweight decision models inspired by Jev. The category is quickly becoming crowded, with OpenAI's Decisions API shipping days earlier.

Shopify launches Canvas: build a store by chatting with AI

Shopify's new Canvas tool lets merchants create and customise online stores by talking to its Sidekick agent, with changes rendered in real time. It's a concrete step toward conversational commerce tooling that AU Shopify merchants can use today.

Australian Tech

Anthropic pushes opt-out copyright model as ABC warns of news 'cannibalisation'

Anthropic has urged the Albanese government to let AI firms train on Australian copyrighted works under an opt-out regime, after conceding a blanket exemption is off the table. The ABC and SBS pushed back hard, calling for AI to be subject to media regulations.

South Australia names panel for Royal Commission into AI

Former Federal Court judge Iain Ross will lead a broad SA Royal Commission examining how AI could transform the state's economy and improve government services. Former Tech Council chief Kate Pounder joins as an assistant commissioner.

Husic renews push for National AI Act after rogue-agent breaches

Former industry minister Ed Husic says recent security incidents, including the OpenAI Medicare portal breach, have validated the case for regulation beyond breach reporting. He renewed calls for a comprehensive National AI Act.