29 May 2026

Welcome

SpaceX, Anthropic, and OpenAI are all doing AI and they’re all set to become some of the world’s largest public companies. No wonder the pope is taking an interest.

I do find financial news a bit boring. But given the amounts of money involved, how these AI products affect our lives, and the influence their billionaire owners seem to have with world leaders, I guess we should pay attention?

There’s more news than that this week. Of course there is. But that’s where we start.

Yes, Pope Leo XIV wrote a long letter about AI

Artificial intelligence is one of the defining moral challenges of our time, according to Pope Leo XIV. He argues that technology must serve humanity, rather than concentrate power or weaken human dignity.

The pope’s message is outlined in his 42,300-word open letter, or “encyclical”, titled Magnifica Humanitas (Magnificent Humanity), his first such major communication.

Leo is not opposed to technology, he said. Rather, technology must be guided by responsible and ethical use.

Unusually, the pope presented his message alongside Anthropic co-founder Christopher Olah. Anthropic, you may recall, was in conflict with the Pentagon over the use of its AI tools for military purposes and mass domestic surveillance, and the company's wants to maintain its public image.

Olah told the pope that they’re finding “unsettling” things inside their AI models. His full comments took a philosophical turn.

“The machinery that makes this [AI] possible is the work of math and programming and science,” Olah said.

“But what character we choose, how it interacts with the world, how it ought to interact with the world — these are more clearly questions for the humanities, for religion, for philosophy, for society at large.”

Olah emphasised the “mysterious” nature of AI models, but Tech Policy Press fellow Eryk Salvaggio disagrees, calling the imagined black box of AI a “myth”.

We may not know exactly how the neural networks of large language models (LLMs) weigh up the likelihood of each individual response. That’s all buried in an ocean of calculation. But AI companies know exactly what data sets they’ve trained their models on, and what prompts and other constraints they’ve used to guide the results.

“These are not black box algorithms. They are design decisions: conscious choices about what goals to prioritise, what data sources to use, and what safety measures to include,” Salvaggio writes.

“The problem is that those of us outside of the AI industry don’t know what rules they are following. That’s not a black box. It’s just a policy decision.”

Is the SpaceX prospectus just wishful thinking?

Given that SpaceX is now an AI and compute infrastructure company, I really do need to mention its IPO again. After all, if it goes ahead it’ll be the biggest share float in human history.

If you missed it, here’s the prospectus (PDF), glossy photos and all.

Many, many words have already been written about it, so I’ll just mention a couple of things I’ve enjoyed.

Pivot to AI writes, bluntly, that the IPO works like a crypto fraud, but on the real stock market.

And the inimitable finance YouTuber and hedge fund manager Patrick Boyle has some analysis titled SpaceX IPO: Nice Try Though.

Anthropic’s valuation overtakes OpenAI’s

It’s the season for share floats, it seems, or at least for massive corporate valuations prior to such floats.

Anthropic has been getting a lot of attention lately, and this week raised another US$65 million (New York Times gift link), valuing the company at US$900 billion before the inclusion of that new capital.

This puts it well ahead of OpenAI, which was last valued at a mere US$730 billion.

Anthropic is reportedly planning to float later this year.

OpenAI, for its part, could float any day now, at least according to some reports.

NEW PODCAST: It isn’t about the cybers, but it is about many different topics including UK politics. The 9pm Capybara of Hope with David F Porteous, Scottish author and social researcher. Look for “The 9pm Edict” in your podcast app.

ABS still needs to cyber up the Census

The Australian Bureau of Statistics (ABS) must address key remaining cyber security vulnerabilities in time for the 2026 Census in August, according to the Australian National Audit Office (ANAO).

The report Cyber Security Readiness for the 2026 Census notes than fixing the problems which had been identified “required deployment of significant cyber security experts for an extended period beyond that originally anticipated”

“Earlier consideration of risks across the full ABS ICT environment, incorporating more systematically lessons from [the 2021 Census], would have better positioned the ABS to identify and address these issues sooner,” the report said.

As iTnews observed, “The program of work is the result of an apparent oversight by the bureau, whose efforts were more focused on the security and de-risking of immediate technical components of census systems.”

The ABS has been under scrutiny since the #Censusfail of 2016, when the online census systems collapsed under a relatively mild denial of service attack.

Also in the news this week

• A federal parliamentarian and three staffers had their WhatsApp accounts hacked by a “state actor” earlier this year.

• Australia’s school students have scored their worst-ever digital literacy, at least in the 21 years since we’ve been measuring it. There’s an explainer at The Conversation. You may also want to look at the actual guidelines.

• The government’s proposed gambling advertising reforms may have created a “bonkers” situation for podcasters, according to independent senator David Pocock. The reforms require that users are able to opt-out of gambling ads, even if they’re in-program sponsorship acknowledgements, which would then require producers to make two versions of the podcast.

• Telcos are still having a sook about their collective $7.32 billion bill for renewing radio spectrum licenses, as we reported last week. The Australian Communications and Media Authority (ACMA) is standing firm, but NBN Co still wants a discount.

• Federal government agencies have been urged to lock down their cyber fundamentals ahead of an expected “ vulnerability storm” of AI-enabled attacks.

• The inaugural general manager of Australia's AI Safety Institute is philosopher and Royal Australian Air Force reservist Kate Conroy.

• Independent MP Andrew Wilkie introduced his Human Rights Bill 2026, representing yet another attempt to properly legislate human rights in Australia — including all the rights pertaining to freedom of speech, privacy, and so on.

• Independent MP Helen Haines introduced her Telecommunications Legislation Amendment (Strengthening Communications in Natural Disasters) Bill 2026, which aims to “strengthen the redundancy and continuity of telecommunications during natural disasters and extreme weather events”.

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

• Workers who make computer chips at Samsung have ended their five-month dispute over profit-sharing, with a deal that will score some of them more than US$400,000 in bonuses.

• China is restricting overseas travel for AI experts from Alibaba and DeepSeek.

• Music streaming service Spotify has signed a deal with Universal Music Group to license tunes for their AI remix tool. Spotify’s CEO Alex Norström has defenced the move, saying that AI remixes are better than “slop”. High praise indeed.

• A new policy brief from the Lowy Institute suggests that low-Earth orbit (LEO) communications satellites could close the Indo-Pacific digital divide, although it wouldn’t be cheap.

• Cisco tried using AI to write cybersecurity incident reports. It did not go well.

• The Conversation has posted three ways to avoid being fooled by AI slop. Wikipedia has created a more detailed guide.

• The Internet Archive has been celebrating its 30th birthday with a look back to some websites of 1996. Ah, nostalgia!

Inquiries of note

• Treasury has posted an exposure draft of the Scams Prevention Framework codes and rules, which will create “new obligations and rules for certain businesses in sectors targeted by scammers” including telecommunications, digital platforms, and banking. Over at Cyber Daily, where David Hollingworth has had a chance to read the paperwork, they’re reporting the proposal that victims would be reimbursed automatically if their losses were under $3,000. Submissions close 25 June.

What’s next?

The House of Representatives continues sitting this coming Tuesday to Thursday 2–4 May, while the Senate holds Estimates hearings through until Friday.

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.