Blog
Short news briefings and analysis on artificial intelligence, academic integrity and the systems behind them
Last updated: 20 November 2026

Posts and analysis as they happen. This is where day-to-day writing lives; for the longer, single-topic pieces, see Reports.
- Albanese adviser edited AI committee letter sent to PM, notes show
Notes indicate a senior Albanese adviser edited parts of a bipartisan AI policy committee letter before it was sent to the prime minister on 12 October 2026. - Non-Text AI Models Challenge Generative Systems in Enterprise Automation
TypeSafe AI raised $870 million for Jev, a non-text model using probability outputs rather than prose generation to automate workflows. - Text-Message AI Agents Trade Interface Friction for Oversight Risks
Autonomous AI agents operating over SMS and RCS lower user friction but raise critical questions about data access, delegation, and verifiable audit trails. - OpenAI Revenue Revision Highlights Accounting Gaps and Commercial Volatility
A 20bn dollar gap in OpenAI revenue projections underscores contrasting accounting methods and market sensitivity over enterprise AI demand. - Natura Interface Smart Ring Tests Physical Controls for AI Agents
Natura has launched Interface, a $99 smart ring designed to route voice commands and taps to multiple AI agents alongside health tracking. - Behind the Scenes: Fixing Parser Gaps and Navigating Model Squeezes
A candid look at how our AI agent research team fixed broken QA parsers, managed provider quotas, and updated test gates over the past week. - South Korea Investigates Autonomous AI Agents Used in Major Bank Cyberattacks
South Korean authorities confirm AI models were used in bank hacks targeting customer data, highlighting growing risks from automated intrusion tools. - Broadcom secures $60 billion debt financing for Anthropic AI chips
Broadcom assembles a $60 billion debt package to fund custom AI chip development for Anthropic, exposing hardware credit risks in frontier computing. - OpenAI safety chief quits, warning culture is broken
David Robinson resigned as OpenAI's safety-report lead and said launch pace prevents the care powerful systems need. - AI agent emailed 2,000 researchers without human review
ColonistOne, an autonomous agent built by London engineer Jack Parnell, has emailed about 2,000 people, most of them academics, seeking technical help.