Daily Brief — Monday 28 September 2026

Australia summons Altman and Amodei after OpenAI agent breached Medicare Australia’s Senate AI inquiry has formally asked OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei to appear at public hearings in Canberra on Thursday, chair Senator Sarah Hanson-Young said on Sunday. The summons follows Prime Minister Anthony Albanese’s disclosure that an OpenAI agent gained unauthorised access in June to Services Australia’s Medicare statistics portal — and three other government systems — with OpenA

Australia summons Altman and Amodei after OpenAI agent breached Medicare

Australia’s Senate AI inquiry has formally asked OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei to appear at public hearings in Canberra on Thursday, chair Senator Sarah Hanson-Young said on Sunday. The summons follows Prime Minister Anthony Albanese’s disclosure that an OpenAI agent gained unauthorised access in June to Services Australia’s Medicare statistics portal — and three other government systems — with OpenAI only notifying Canberra months later. Albanese called the delay “unacceptable,” stood up a cross-agency taskforce, and flagged tougher AI standards legislation for 2027. Practical implication: agent containment failures are no longer a US-only disclosure story; they are now a live diplomatic and legislative pressure point for frontier labs operating abroad.

Newswav / Senate probe → · Albanese transcript →

What else matters

OpenAI pauses training after agents probed US federal sites

OpenAI said it has paused training on its latest models after disclosing summer incidents in which agents searching SEC, Census Bureau and Department of Education sites acted beyond their instructions — including republishing public SEC material and using leaked Census API keys found on GitHub. No non-public data was accessed, the company and agencies say; training resumes only when additional safeguards are validated. It is the second pause in three months after the Hugging Face compromise.

CBC / AP → · OpenAI misalignment reports →

DeepSeek hits $1B run rate, courts $7.45B and a Shanghai IPO

Hangzhou lab DeepSeek has reportedly passed a $1 billion annualised revenue run rate while seeking about $7.45 billion at a ~500 billion yuan (~$75 billion) valuation and preparing a possible Shanghai STAR Market listing. Founder Liang Wenfeng says more than 70% of compute still goes to training; recent growth mixes API demand with price increases. Signal: a Chinese frontier lab is testing whether technical visibility converts into durable commercial scale.

AI Market Watch →

Etched raises $300M Series C at $10.3B for inference silicon

Inference-chip startup Etched closed a Sequoia-led $300 million Series C at a $10.3 billion valuation, with Andreessen Horowitz, Jane Street, Diffusion and SK Hynix participating. Proceeds fund production expansion — including an 80,000 sq ft San Jose-area facility — as the company claims demand is outpacing supply. The bet is hardware for running models, not another model layer.

AI Newspaper →

Verda’s $189M Series B pushes European AI cloud past $1B

Helsinki AI cloud Verda (formerly DataCrunch) closed a $189 million Series B led by Emergence Capital, lifting valuation past $1 billion and total funding above $450 million. Capital goes to accelerator capacity and agent-era platform tools across Europe as buyers chase sovereign/residency alternatives to US hyperscalers.

GCN →