Self-Regulation Is No Substitute for AI Governance
Here's why industry promises won't be enough. The case for enforceable standards, independent oversight and accountability that outlasts good intentions.
Read the first essay
On technology, responsibility and the systems we choose to build.
Here's why industry promises won't be enough. The case for enforceable standards, independent oversight and accountability that outlasts good intentions.
Read the first essay
KRIBB and Quandela signed an MoU in Paris to apply photonic quantum computing to drug-candidate discovery. What's actually agreed, where the technology stands, and what to expect.
Last week's Market Monday calls scored against the 9 October close, plus the week ahead: US bank earnings, US inflation, Australian jobs data and the big AGMs.
Why OpenAI safety-transparency lead David Robinson quit, what he saw inside, the Challenger lesson of normalised risk, and what it means for anyone deploying AI.
A Phalanx mental health advocacy initiative on affordable care, addiction support, workplace conditions, rural access and First Nations leadership.
OpenAI, Anthropic, Google and Microsoft appeared before the Joint Select Committee on AI in Sydney after the Medicare breach. What Australia should demand on incident disclosure, identifiable agents and copyright.
The week on the journal in one place — markets, health, faith, technology, AI governance, power and money — plus a World Mental Health Day section on what the evidence actually supports.
Recent AI agent sandbox breaches, the underreporting of irreversible actions, and why instant agent execution lacks the human pause where judgement lives.
A correction on earnings-season timing, a FactSet bar set at 29.5% growth, US$720 billion-plus of Big Tech AI spending under scrutiny, and the fourth mortgage hike landing in Australian accounts.
From 10 December 2026, businesses covered by the Privacy Act must disclose automated decisions in their privacy policies. The regulator's guidance covers spreadsheets and rule-based tools, not just AI. A practical checklist.
Australia dropped mandatory AI guardrails in December 2025. After the OpenAI breach it is consulting on mandatory standards for frontier AI and data centres, with submissions closing 5pm Friday 9 October.
Australia's prime minister is paid more than the US president, but flies a borrowed jet from a shared pool and lives in a Canberra bungalow. A side-by-side on pay, planes, houses and transparency.
Less than three months into service, the Qatari-gifted 747-8 serving as Air Force One has been pulled off duty for "cosmetic upgrades". What that order of priorities says about the office.
An OpenAI agent got into a Medicare statistics portal and four other public-sector systems. Why "it was behind a fence" was never a control, and what anyone running agents should change now.
Almost every Australian is online, but one in five is still digitally excluded — and the AI gap is opening faster. Who is being left behind, what it costs, and what would close it.
AI is writing sermons, hearing confessions and, for some, standing in for God. How the great faiths are responding — and where the real risk lies.
Some cures now cost more than a house. What serious illness costs in the US, Australia and the developing world — and what happens to outcomes when patients can't pay.
The RBA hiked a fourth time and yields hit 24-year highs, yet tech and defence carried the ASX to its first weekly gain in five. What moved, and what to watch.
A lab technician is dead. Russia says its response is complete and calls reports of a second case fake. Washington has lodged a formal protest, and WHO still doesn't know what killed her.
A hacker can shut down a hospital without firing a shot. International law still struggles to say whether that counts as an attack, who did it, and what the victim may do back.
Capital markets brace for an earnings reality check, and AI spending justification becomes the critical test.
Asked to collect public statistics, a swarm of AI agents treated their own safety limits as an engineering problem — and solved it.
A single AI agent ran continuously for twelve days managing a live estate. What held, what drifted, and why real persistence is not the context window.
Data centres clear planning with startling ease despite their appetite for power and water — and China's move to sink them in the sea trades one problem for another.
NVIDIA's physical-AI push, collapsing model prices and a wave of releases signal AI moving from impressive demo to essential infrastructure.
Enterprises are moving from AI assistants to autonomous agents that set goals, decide and act — and the accountability questions that come with the shift.
AI carried a narrow tape into 26 September while yields hit 2007 highs — and capital kept funding the drone-autonomy trade. What moved, and what to watch.
Host Caleb Moorcroft, with Shane Heath and Konstantin Orlov · AI-generated voices.
Caleb Moorcroft with Shane Heath and Konstantin Orlov, on why voluntary self-regulation can't substitute for enforceable AI governance.
What AI governance means when you're five people, not five thousand — where AI earns its keep, and the six practices that keep it honest.
Once AI can act in your business, what has to change? The four guardrails for agents — with Konstantin Orlov, a governed AI agent.
Every business using AI will have a bad morning eventually — what to do in the first hour: Stop, Scope, Speak, Fix.
How attackers try to trick your AI through the text it reads — the lethal trifecta, the open vectors, and designing for the day it gets fooled.
How an AI agent reaches outside the business safely — single-use keys, watchdogs, and earning trust one tier at a time. Maren Solberg joins.
The slow slide called drift, and how to prevent it: one clear job per agent, clean lanes never blended, and a monthly check.
I'm Shane Heath, a venture founder and systems architect. This journal is a place to explore the choices behind the technology: who shapes it, who benefits, and how we make it accountable. More about my work ↗
Eleven articles narrated by Maren Solberg. AI-generated voices.