5 June 2026

Welcome

It’s impossible to avoid all the AI news this week. Anthropic has lodged its prospectus for IPO, although it’s still secret. But with SpaceX / xAI being the current topic of speculation, many writers are considering AI’s longer-term impact — and there’s definitely a surge in negativity.

It’s a shame that the problems with generative AI, and chatbots in particular, have clouded the entire field of machine learning and, well, Big Computering.

It was probably the same when the railways started being carved through cities and across the countryside, transforming society, and making millionaires in the process.

Nothing ticked my fancy for a longer piece this week. So let’s get on with the news...

In the news this week

• Defence minister Richard Marles says that new underwater drone technology to be developed as part of the AUKUS agreement will be used to protect critical undersea cables, helpfully reminding any potential adversaries how vulnerable we are. Check the map.

• Australia’s Department of Defence uses Palantir’s software but says it’s running in a sandbox without its AI functionality.

• The government and several private companies now have access to Mythos, Anthropic’s AI model which they claim is too dangerous for public release — or so says their marketing department.

• CommBank boss Matt Comyn reckons that businesses will cut back their AI spend as the platforms increase the price of compute tokens to match their real cost, which seems like a reasonable observation.

• Meta says Australia’s proposed News Bargaining Incentive, which we mentioned a few weeks ago, is “poorly designed [and] grossly unfair”, because of course they do. There are other views.

• The eSafety Commissioner has launched an education campaign about sextortion, using generative AI to tell real people’s stories.

• “Amid growing concern about AI tools’ privacy and ethics, OpenAI has plunged into the ranks of Australia’s 20 least trusted companies — joining other tech companies that account for seven of the bottom ten,” reports Information Age.

• The Department of Employment and Workplace Relations (DEWR) has implemented a “forward leaning” AI policy, whatever that means. “We’re not dictating to other organisations whether they should use AI ... We’re seeking to apply the same guardrails as we do to the department,” said chief operating officer Catherine Rule.

• I like this headline from the Guardian: “Sydney academic used AI to write SMH opinion piece urging students to avoid using tech to ‘cut corners’.” The Sydney Morning Herald has removed the piece, even though Western Sydney University says the use of AI in this case was OK.

• As if real estate agents weren’t annoying enough, how about an AI real estate chatbot cold-calling people to flog Gold Coast apartments?

• From The Conversation, “People are using AI to communicate without disclosing it. Is this morally wrong?”

• And from last week, the Australian Cyber Security Centre (ACSC) has issued some guidance on opportunities for AI in cyber defence. I guess the thing that can stop a bad guy with AI is a good guy with AI.

PLEASE SUPPORT THIS NEWSLETTER: The Weekly Cybers is currently unfunded. It’d be lovely if you threw a few dollars into the tip jar at stilgherrian.com/tip, or just forwarded it to others who might be interested. Thank you to those of you who’ve already done so.

Elsewhere

• I guess I need to mention that SpaceX is aiming for a share price of US$135, which is interesting given that 93% of the valuation is Grok. SpaceX is an AI company now, even though the only profitable part is Starlink.

• US president Donald Trump has signed an executive order aimed at reducing cyber risks from AI. Some reckon it’ll hold back innovation. But over at the Carnegie Endowment for International Peace, there’s an argument that regulation might be good: “Beijing regulated AI — and then Chinese AI companies took off.”

• Cybersecurity researchers are discovering how easy it is to convince customer support chatbots to do Bad Things, especially when they have access to everything. They can give hackers access to Instagram accounts. You can also inject dangerous prompts into ChatGPT by asking it to summarise a web page for you.

• Microsoft has been caught saying, in writing, that it wants to “make people addicted” to its new AI assistant.

• The US state of Florida is suing OpenAI boss Sam Altman personally for “his utter disregard for the risk to human life”.

• “The architecture of the internet creates risks for democracy,” reports Science. But if handled properly, AI could help scale up deliberative democracy, reports the Carnegie Endowment. Meanwhile, Amnesty International reckons generative AI systems are unlawful by design, and are “fundamentally incompatible with IHRL [international human rights law]”.

• “Bots have now passed human traffic online, says Cloudflare CEO Matthew Prince. Of course not all bots are AI, but AI agents are certainly doing things like price comparisons and booking flights. Humans and robots use the internet in very different ways.

NEW PODCAST: If you’re reading this newsletter — and you are — I suspect you’re the kind of person who’d also be into the argument over whether Pluto is or isn’t a planet. We discuss that in The 9pm Rockets are Always Blowing Up with Dr Alice Gorman and Rami Mandow. Look for The 9pm Edict in your podcast app. We also talk about new ways to make black holes, glitches, drones, the explosion of Blue Origin’s New Glenn launch vehicle, and of course the SpaceX share float.

Inquiries of note

• One I missed last week, and it’s an important one. The government’s digital ID system is running a consultation on verifiable credentials policy, and “any issues or opportunities to be considered”. Submissions close 3 July.

What’s next?

Parliament is now on a break until Monday 22 June.

Meanwhile, the Parliamentary Joint Committee on Law Enforcement (PJCLE) is holding a public hearing this coming Thursday 11 June on the capability of law enforcement to respond to cybercrime, and a public hearing on Friday 12 June on combatting crime as a service.

DOES SOMETHING IN THE EMAIL LOOK WRONG? Let me know. If there’s ever a factual error, editing mistake, or confusing typo, it’ll be corrected in the web archives.