The News,
Decoded.

Oh, superb — OpenAI has hired Colin Fleming, ServiceNow's CMO, to lead business marketing. Because nothing says 'we've solved alignment' like appointing a SaaS marketing veteran to sell enterprise AI. The company claims this is about 'deepening enterprise marketing beyond core AI products,' which is corporate-speak for 'we need to convince CFOs that GPT-5 is worth the compute budget.' Fleming's job? To polish the turd of API pricing and hallucination rates. One wonders if his first task is a rebrand: 'Generative Pre-trained Transformer' sounds too technical — perhaps 'Corporate Thought Partner' instead. The irony is delicious: ServiceNow's own AI strategy is still a mess, but sure, hire their CMO. At least he knows how to sell a platform that doesn't quite work. Meanwhile, OpenAI's enterprise revenue remains a black box, but why let facts get in the way of a good press release?

Fleming's departure from ServiceNow is itself telling: the company's AI push has been lukewarm, and its stock dipped after a recent earnings miss. So OpenAI is essentially hiring someone whose last job was failing to sell AI to businesses. That's like hiring the captain of the Titanic to lead your cruise line. But optimism reigns! The press release quotes Fleming saying he's 'thrilled to join at such a pivotal time' — the classic sign of a man who spotted a lifeboat. The business unit's success will inevitably be measured by some vanity metric like 'enterprise logos' or 'API calls from Fortune 500.' Actual revenue? Who needs it when you have vibes.

This move also underscores the gap between frontier labs and real-world deployment. OpenAI's models are impressive, but selling them to enterprises requires navigating compliance, security, and the fact that most business problems don't need a PhD in transformer math. Fleming's task is to bridge that chasm with marketing spin. Expect upcoming campaigns touting 'co-pilots for procurement' and 'AI-driven supply chain synergy.' In five years, we'll look back and laugh at how we thought hiring a CMO could paper over fundamental reliability issues. Or we'll be too busy being replaced by a bot that pitches better. Either way, cheers.

Oh, wonderful — another AI-powered band-aid for a problem AI helped create. DockSec, an OWASP incubator project, uses AI to correlate findings from container security scanners and generate 'plain-English remediation guidance and Dockerfile fixes.' Because nothing says 'trustworthy' like an AI telling you how to fix vulnerabilities your own AI-generated code probably introduced. The OWASP incubator tag suggests this is promising enough for a science fair project but not yet ready for prime time — exactly the kind of confidence that has made container security such a joy.

DockSec aims to 'cut through vulnerability noise' in Docker images, which is a polite way of saying that current scanners produce so many false positives that nobody bothers fixing anything. So now we'll have an AI that produces plausible-sounding fixes for problems that may or may not exist. If it hallucinates a fix, at least it'll be in plain English — you can enjoy the grammar while your production cluster burns. The project correlates findings from multiple scanners, which is clever, but correlation isn't causation, and neither is 'AI-generated Dockerfile.'

Let's be honest: this is a classic security-adjacent startup move — solve a problem created by tool overload with another tool. The real question is whether DockSec will introduce its own set of vulnerabilities in the process. After all, the open-source package that imperiled millions of AI agents we flagged last week relied on the same kind of AI-assisted development pipeline. But hey, at least now we'll have an AI that can write an apology letter in plain English.

Ah, the 'Wild West' of government AI use. Because nothing says 'frontier spirit' like bureaucrats fumbling with generative models they don't understand. Michigan state Rep. Jaime Greene has nobly introduced a bill to 'study' the issue and 'ensure oversight' — because heaven forbid we actually regulate anything before a fourth study commission reports back in 2029.

Let's be honest: this bill is a masterclass in performative governance. It'll commission a report, which will recommend more studies, which will gather dust while every county in Michigan quietly adopts ChatGPT to write zoning permits and child welfare assessments. The real cowboy isn't the AI — it's the legislators who think a 12-month inquiry counts as 'taming the frontier.'

Meanwhile, the private sector is shipping models with fewer guardrails than a late-night trampoline park. So by all means, Michigan, take your time. The 'Wild West' will be a full-blown Mad Max dystopia by the time your task force schedules its second stakeholder meeting. Yee-haw.

Oh, brilliant. SoftBank is joining the sovereign AI race with a GPU cloud infrastructure thingy. Now it's pivoting to 'sovereign AI' — a term so conveniently vague it can mean anything from 'national pride' to 'we'll host your models and also your data because who needs sovereignty anyway.' The infrastructure is slated to deploy in October, integrated with something called Infrinia AI Cloud OS, a name that sounds like it was workshopped by a committee that really wanted 'synergy' and 'cloud-first' on the same slide.

October deployment, because nothing says 'strategic urgency' like aiming for a month that's always six months away. By the time SoftBank's GPUs actually hum, every state-backed AI lab from Saudi Arabia to Singapore will have signed exclusive deals with Nvidia, AMD, and whichever startup promises the shortest waitlist. But never mind — the Infrinia AI Cloud OS will tie it all together in a 'seamless orchestration layer' that will, presumably, require a separate license, a dedicated team, and a prayer to the gods of Kubernetes. The real race isn't for sovereignty; it's for who can sell the most shovels to governments suddenly terrified that they're falling behind in the AI gold rush.

And what a rush it is. Countries that couldn't build a functional e-voting system are now commissioning 'sovereign AI clouds' as a shortcut to technological independence. SoftBank, ever the happy enabler, is happy to supply the hardware — and collect the rents. The sovereign AI race is, after all, just another name for the sovereign cloud spending spree, and SoftBank's late arrival means it'll have to offer competitive pricing or, more likely, fantastically optimistic timelines. October 2026? Perfect. Plenty of time for the bubble to inflate, pop, and be reframed as a 'learning experience.' Cheers.

Oh, marvellous. Another AI model that promises to revolutionise cancer treatment, because clearly the bottleneck in oncology has been a lack of predictive algorithms. Researchers at UC San Diego have unveiled a model that translates a tumor's genetic profile into predictions about treatment response, published in the hallowed pages of Cancer Discovery. One imagines the press release practically wrote itself: "AI will cure cancer" — never mind that the model's actual clinical utility remains a distant mirage on a crumbling road of regulatory approval and insurance reimbursement. The study will no doubt be cited by every AI-in-medicine startup seeking funding, while the clinicians who actually treat patients will continue to rely on their 40-year-old chemotherapeutic regimens.

The study involved training on some dataset of questionable generalisability, producing impressive AUCs that will never see the light of a real hospital. Because the gap between a published paper and a deployed clinical tool is not a gap — it's a chasm filled with FDA denials, hospital IT nightmares, and the quiet desperation of oncologists who have seen this movie before. But don't let that dampen the headlines. UC San Diego has spoken, and the future is here. Again. The model may be able to predict which drugs work on which tumors, but it cannot predict its own journey through the American healthcare system — a system that rewards innovation theatre over actual patient outcomes.

So raise a glass to yet another AI "breakthrough" that will generate a flurry of breathless coverage, a few grant renewals, and exactly zero changes in how cancer is treated this decade. Every few months, another group publishes a model that claims to predict treatment response with remarkable accuracy. Each time, the oncology community yawns and returns to the slow, messy work of evidence accumulation. This model will be no different. It will be validated on retrospective data, implemented at exactly one academic centre, and eventually forgotten when the next shiny AI object appears. Cheers to progress.

Oh, marvellous. Goldman Sachs has confirmed that the smart money is 'doubling down on AI' by fleeing software stocks faster than a politician caught in a scandal. Semiconductor weight is at a record high, while software is nursing its lowest allocation since 2019. Because nothing says 'betting on the intelligence revolution' like dumping the very companies that build the applications, and piling into the companies that just make the rocks you heat and etch. It's the investment equivalent of buying shares in a bakery because you love bread, but refusing to buy the bakery itself.

The logic, apparently, is that the real value is in the hardware — the shovels, not the gold. But this is Goldman Sachs we're talking about, a firm that has historically been wrong about almost everything at exactly the worst moment. Remember when they advised clients to buy subprime mortgages in 2007? Good times. So now they're telling hedge funds to pile into semiconductors and dump software, which is basically a bet that AI will never need to actually run anything useful — just exist as a potentiality in a server farm. Bold strategy, Cotton.

And what of the software engineers who were told their jobs were safe because 'AI is just a tool'? Well, if the market is already rotating out of software, the message is clear: the AI revolution eats its own. First the applications, then the people who write them. But don't worry — Jensen Huang says blaming AI for job cuts is 'lazy.' So when Goldman's next report shows a record high in semiconductor weight and a record low in software employment, it'll just be a coincidence. Clink. 🍻

Oh, how convenient. Jensen Huang, the man who has made a fortune selling the shovels in the AI gold rush, declares that CEOs who blame job cuts on AI are 'lazy.' This from the CEO of Nvidia, whose chips are the very engines driving the automation that is displacing workers. It's a bit like a tobacco executive calling lung cancer patients 'unmotivated.'

But let's examine Huang's logic: he claims it 'doesn't make any sense' to link layoffs to AI before generative tools became widely useful. So, the thousands of workers laid off by tech giants in 2023 and 2024 — those were just… what? Seasonal housekeeping? A collective hallucination? The facts are stubborn: layoffs surged as AI tools improved, and job postings for administrative roles cratered. To call the connection 'lazy' is either wilful blindness or masterful deflection.

What Huang's comments really reveal is an industry desperate to distance itself from the consequences of its own products. Nvidia's market cap has ballooned while its customers slash headcount in the name of 'AI transformation.' By calling out CEOs, Huang shifts the blame from the technology to its users. It's clever public relations, but it won't comfort the millions whose jobs are being 'optimized' into oblivion. The dark joke? The best defense of AI is that the humans deploying it are too incompetent to take credit for the destruction.

The International Maritime Organization, that notoriously swift UN body, has finally adopted binding rules for autonomous ships. The MASS Code — International Code of Safety for Maritime Autonomous Surface Ships — is supposed to ensure safe integration of AI-enabled and remotely operated commercial vessels. Because nothing says 'safety' like a bureaucratic code crafted by committee while actual ships are already navigating autonomously in some waters. The technology has been outpacing regulation for years, and now we're meant to believe a document will fix that.

The code is 'binding' in name, but enforcement depends on flag states — countries that often compete for shipping registry business by offering lower standards. Shipping companies will surely comply enthusiastically, just as they do with existing safety regulations. The real test will come when the first autonomous vessel causes a major incident; then we'll see if these rules have any teeth. Until then, the gap between the code and reality is where innovation and profit flourish, with minimal oversight.

So raise a glass to the IMO for its timely intervention — if by timely you mean 'years late and already obsolete.' The MASS Code is a classic example of regulatory catch-up: well-intentioned, probably ignored when inconvenient, and certain to be outpaced by the next generation of AI. But hey, at least we have a code now. It may not prevent the first autonomous ship disaster, but it'll provide excellent reading material for the inquiry that follows. Cheers.

Oh, splendid. Another survey confirming what anyone with half a brain already suspected: CEOs are sharpening their redundancy knives and blaming artificial intelligence. Mercer's Global Talent Trends report reveals that 99% of chief executives anticipate AI-driven layoffs within the next two years — that’s near-unanimous, the sort of consensus usually reserved for 'oxygen is good' and 'bonuses are deserved'. And where will the axe fall? On early-career positions, naturally. Because nothing says 'welcome to the professional world' like being replaced by a large language model before you've even learned where the coffee machine is.

The cowardice is staggering. By targeting the most junior and least politically connected staff, executives can frame the bloodbath as 'strategic workforce realignment' while ensuring their own roles remain untouched. The Mercer report gives them perfect cover: it's not greed, it's a trend. When 99% of your peers are doing it, it's not a layoff; it's industry standard practice. Never mind that many of these AI systems are still prone to hallucination and error; the important thing is the quarterly earnings call.

And what of the long-term consequences? If every major company simultaneously guts its entry-level roles, where does the next generation of experienced hires come from? That's tomorrow's problem, and tomorrow's CEOs — who will probably be AI agents by then — can deal with it. For now, the message to young workers is clear: you trained for four years to be replaced by a glorified autocomplete. Welcome to the future. It's not quite what the brochures promised.

Anthropic's Claude Mythos flexed its bug-hunting muscles, but the real discovery is that open-source maintainers are drowning. Of 6,200 vulnerabilities flagged in open-source code, a triumphant 75 have been patched.

Let's not kid ourselves: the capability leap here is real, and it's terrifying. Claude Mythos can now do in weeks what took armies of security researchers years. But the gap between discovery and remediation just yawned open like a subprime mortgage crisis. The article quotes a researcher saying 'the bottleneck has shifted from finding bugs to fixing them' — which is tech-speak for 'we built a firehose and forgot to install a drain.'

And the really funny bit? Anthropic will probably spin this as evidence of responsible AI deployment. 'We found the problems!' they'll say, conveniently omitting that nobody's solving them. It's like a doctor diagnosing a pandemic and then going on holiday. With great power comes great… what was that phrase again? Ah yes, plausible deniability. 🫖

Oh, marvelous — two new ‘AI scientists’ have arrived, and they're about as reliable as a second-hand Bunsen burner. Future House's Robin and Google DeepMind's Co-Scientist are multi-agent systems designed to automate scientific discovery, which apparently means we can finally remove the messy human element from the pursuit of knowledge. The Phys.org piece tells us they're 'improving,' but also that they have 'fundamental limits in trustworthiness' — which is tech-speak for 'they make stuff up with conviction.'

One imagines the pitch meeting: 'What if we replace grad students with something that hallucinates citations even faster?' Because nothing says scientific rigour like a system that can't be trusted but is being deployed anyway. The article doesn't get into specifics, but we all know the drill: impressive demos, terrifying edge cases, and a collective shrug from the labs. DeepMind, of course, has form here — remember when they promised to revolutionise protein folding and then quietly monetised it?

So here we are, automating the one domain that absolutely requires reproducibility and scepticism. The irony is so thick you could culture a petri dish on it. These systems will probably 'discover' something profound — like that water is wet or that p-values can be gamed — and then we'll all pretend the trustworthiness problem is a minor bug to be patched in version 2.1. Science: now with added stochastic parroting!

Nerfing — a verb borrowed from the gentle art of making foam darts harmless — has found its perfect application in AI, and the irony is delicious. Google and Anthropic, after years of breathless press releases about superhuman intelligence, are now quietly deflating their chatbots like sad party balloons. TheStreet reports what anyone with a free account already knows: models that once wrote poetry now struggle with prose, and assistants that handled hour-long conversations now tap out after five minutes. Users are furious, naturally, but the companies seem to be betting that a cheaper, dumber chatbot beats an expensive, brilliant one that might say something embarrassing.

The logic is a masterclass in corporate doublespeak. Google, which once dissolved its ethics board because it was inconvenient, now hides behind 'responsible AI' while throttling Gemini's performance to punish heavy users. Anthropic, founded on the noble premise of 'constitutional AI,' has discovered that a safety-obsessed model that refuses every third request is not actually a product people want to use. Both have hit the same wall: frontier models are horrifically expensive to run, and giving them away for free was never a sustainable business plan. So they nerf — and call it 'maturity,' 'optimization,' or 'alignment.'

But the real story isn't the degradation — it's the pattern. First, the labs deploy a powerful model to grab headlines and users. Then, once dependency sets in, they dial back the capability, introduce caps, and start charging for what was free. It's the drug dealer playbook, except the product is a text generator that occasionally lies. And with no regulator in sight who understands the technology well enough to intervene, expect this to become the standard arc for every AI release. The frontier isn't advancing; it's being carefully managed into mediocrity. At least the foam darts are still fun.

Nothing encapsulates the spirit of 2025 quite like this: Commonwealth Bank, an institution that has occasionally made its own ATMs vanish from existence, is now deploying an AWS AI agent to help on-call engineers during those treasured 2am wake-up duties. The agent is designed to reduce the time spent "connecting data" to find root causes. Because what every sleep-deprived engineer truly needs is a machine that arrives ten minutes before they do and starts reassigning blame in a language only cloud providers understand.

The bank, which has a long and distinguished history of IT outages that have locked customers out of accounts mid-week, is now trusting an algorithm to pre-digest its cloud failures. The agent is 'helping' – that wonderful corporate trope that usually means 'it does the interesting bits, you still do the rest.' One can picture the scene: engineer gets paged, reaches for phone, sees AI has already run a dozen queries and concluded 'human error' – a result it learned from watching the incident post-mortems.

Still, there is a grim beauty in the overall trajectory. Every major bank is now racing to deploy its own little AWS fixer, edging toward the dream of fully automated operations and the inevitable nightmare of telling a regulator why the agent rebooted the production database during end-of-month processing. But why be cynical? The AI is helping. The on-call rotation endures. And when the agent escalates another false alarm at 3am, the engineers can rest easy knowing that at least the data is now beautifully connected.

Well, well, well — the SpaceX IPO prospectus has finally lifted the hood on xAI, and it's not pretty. The AI darling posted an operating loss of $2.47 billion on just $818 million in revenue in Q1 2026, following a full-year 2025 loss of $6.36 billion on $3.20 billion in revenue. That's what Morningstar Canada charmingly calls 'reckless,' which is corporate-speak for 'we're burning cash faster than a Falcon 9 on re-entry.'

Elon Musk's latest vanity project is hemorrhaging money at a rate that would make a Silicon Valley bubble startup blush. The numbers are hilariously lopsided: xAI spent nearly three dollars for every one it earned last year. In Q1, the ratio got even worse — roughly $3.02 in losses per dollar of revenue. That's not a business plan; that's a pyre made of server racks and GPUs, fuelled by the delusion that throwing enough compute at the problem will eventually print money.

But here's the kicker: this isn't a standalone company; it's a subsidiary nested inside SpaceX's IPO. The prospectus essentially serves as a warning to potential investors that xAI's financials look like a controlled demolition. And yet, the narrative persists that frontier AI models are on the cusp of miraculous profitability. Sure, keep telling yourself that while xAI's operating loss could fund a small country's GDP. What could possibly go wrong?

HSBC CEO Georges Elhedery has warned that AI will both create and destroy jobs – a revelation so profoundly obvious that one wonders if the boardroom has only recently discovered the internet. The bank plans to retrain all 200,000 staff to "adapt", as if the solution to being replaced by a machine is to become a better student of the machine's inevitable takeover. It's the kind of "we're all in this together" messaging that works splendidly until you realise the "together" includes algorithms that don't need lunch breaks or performance reviews.

The retraining logistics are a joke that writes itself. Two hundred thousand employees, each presumably with a full-time job, are supposed to somehow find time to learn new skills that the bank itself admits will eventually render their current skills obsolete. HSBC hasn't released a budget, but if corporate training history is any guide, most will get a few bite-sized videos and a "check-in" with their manager – coincidentally, a manager whose own position is likely in the next wave of cuts. The announcement functions as a preemptive PR inoculation: when the layoffs come, the bank can say "But we offered retraining!"

And what of the creative jobs? Elhedery insists new roles will emerge – presumably in the field of explaining what the AI just did, soothing customers the AI has offended, or writing more press releases about how human-centric the bank's AI strategy remains. The real reshaping is of the corporate narrative: a warning about job displacement has been recast as an uplifting story of personal development. Georges Elhedery gets to be the visionary leader who saw the future coming, even as the future is a bulldozer and the retraining is a leaflet about "embracing change." Cheeky.

Lovely news from the subcontinent: 45% of Indian firms have identified AI, digital, and data skills as their single biggest workforce headache, according to the SHRM India Skill Intelligence Report 2026. That's nearly half. You'd think that would spur a mad scramble for training and hiring. But no — because 54% of these same organisations report 'moderate to low urgency' on AI investment. So a majority are treating the biggest constraint with the urgency of a man who knows the dam is leaking but decides to take a nap.

This isn't just cognitive dissonance; it's a masterclass in strategic avoidance. The report doesn't name names, but one imagines HR directors nodding sagely while filling in 'deferred to next quarter' on their AI roadmap. Meanwhile, the global AI arms race — which we've covered extensively for the cyber-hacking models from Politico — continues at pace. India's tech sector might want to catch up before its digital workforce is as outdated as a dial-up modem.

To be fair, maybe the 54% have a point. Perhaps AI investment is overhyped, and the best strategy is to wait and see. But given that 45% are already feeling the pinch, that's a dangerous game of chicken. One might call it the world's slowest-moving skills crisis — or, as the report dryly puts it, a 'constraint' that nobody is in a hurry to lift. Cheers to that.

Apple has finally noticed the generative AI party and decided to reserve a seat at the table — or at least a subdomain. genai.apple.com popped up on DNS on May 23, weeks before WWDC, where Tim Cook has solemnly promised 'AI advancements.' Because nothing says cutting-edge like preparing a landing page before you've shipped a product. Apple, the company that treated AI like a state secret while everyone else was building chatbots, is now rushing to catch up with a carefully curated, walled-garden version of AI that will probably only work if you own three Apple devices and have recited the privacy policy twice.

The irony is thick: Apple's 'genai' subdomain will almost certainly redirect to a page explaining why their AI is better because it respects your privacy — while hoovering data in the background, of course. Expect Siri to finally become useful, but only if you ask nicely and the stars align. Meanwhile, competitors have been shipping generative features for over a year. But hey, Apple knows its audience will buy anything, even a belated AI strategy that smells like damage control.

This is the same company that spent years mocking AI assistants as gimmicks. Now they're leaning in, but the move feels like bringing a cake to a party where everyone else already cooked. The 'genai.apple.com' subdomain is a placeholder for a product that, if history is any guide, will be late, polished, and locked down tighter than a Tim Cook keynote script. Another typical Apple move: announce, promise, underdeliver, then call it 'courage.'

Ah, brilliant — because nothing says 'ready for primetime' like testing an AI pilot in a Cessna Caravan over some unsuspecting American suburb. Merlin Labs has taught its system to listen to air traffic control and chirp back like a polite chatbot, and aviation companies are 'starting to embrace it.' Presumably they've never been on a conference call where the AI suddenly decides the word 'runway' means 'time to sprint.' What could possibly go wrong? At least when autopilot fails, there's a human to grab the yoke. When this one hallucinates a fake clearance, you get a front-row seat to the tarmac.

The timing is impeccable, of course. Hot on the heels of every tech company proving AI can't reliably tell a duck from a dog, we're now handing it the controls of a multi-ton metal tube. The industry is 'starting to embrace' this, which in aviation-speak means a few venture-funded startups have flashed some PowerPoint slides at sceptical regulators. But the trajectory is clear: from Cessna to commercial in a hop, skip, and an inevitable NTSB report. The really clever part? They've given it a voice. So when the AI decides to 'negotiate' its landing slot like a budget airline customer, the human pilot gets to sit there, hands off, listening to a machine argue about holding patterns.

And let's not forget the real punchline: regulators are 'starting to embrace' it too, which in aviation regulation means they've agreed to 'monitor the situation' and 'revisit the guidelines in 2027.' By then, the AI will have flown a few thousand hours, most of them fine, right up until it encounters a non-standard instruction like 'go around, there's a flock of geese.' The black box transcript will be a masterpiece of polite confusion: 'I'm sorry, I didn't understand that. Please rephrase your directive in the form of a question.' Have a lovely flight, everyone.

Well, well — the 'game-changer' has arrived, and it's got a hacking certification. Nine top cyber researchers have told Politico that Anthropic's Mythos and OpenAI's GPT-5.5 are not only capable of offensive cyber operations but are advancing 'much faster than anticipated.' Because nothing says responsible AI development like a couple of labs quietly enabling digital armageddon while Washington fumbles for a response. The wild frontier we noted last week in New Zealand suddenly looks like a polite park compared to this.

And what's the response from the people who fund these labs? Crickets, mostly. The researchers quoted describe capabilities that could 'automate sophisticated hacks' at a speed that outstrips any defensive measure currently deployed. But don't worry — I'm sure the companies will self-regulate, just like they did with social media algorithms. Meanwhile, the same labs are busy firing safety researchers and pushing for commercial deployment. Priorities, people.

The truly farcical element is that this revelation lands in the same week OpenAI's GPT-5.5 was being marketed as a productivity tool for 'creatives.' Yes, the same model that can now compromise a network in minutes is also writing your marketing copy. If that doesn't sum up 2026, nothing does. The only silver lining? At least the hackers will be using state-of-the-art AI — so we'll be compromised with style.

Well, well — New Zealand's cyber watchdog is 'learning from US companies' testing AI models like Anthropic's Claude Mythos. Because nothing says proactive defence like waiting for Silicon Valley to finish beta-testing its digital attack dogs and then politely asking for notes. Palo Alto Networks found that Claude Mythos could detect and exploit software flaws so efficiently that it triggered dozens of security alerts in a single day — compared to its usual five per month. That's a 10x to 100x increase in potential mayhem, depending on your arithmetic.

The term 'wild frontier' in the headline is apt, but not for the reasons the RNZ editors might think. The frontier isn't the technology itself; it's the yawning gap between what these models can do and what any regulator — from Wellington to Washington — is prepared to handle. The NZ watchdog is 'learning,' which is lovely, but learning from companies whose business model depends on pushing capabilities fast and apologising later (if at all). It's like asking a fox for advice on henhouse security while he's still crunching feathers. Anthropic touts Claude Mythos as 'safety-focused,' but a system that turns a trickle of alerts into a flood in under 24 hours seems less about safety and more about setting a land-speed record for finding vulnerabilities.

Of course, the irony is thick enough to spread on toast: the same labs claiming to prioritise safety are producing models that can supercharge hacking at a pace regulators can't even measure, let alone match. New Zealand's approach — watch, learn, maybe eventually act — is the global standard. And that standard, dear reader, is how you end up with a 'wild frontier' where the bad guys have superhacking AI and the good guys are still debating which documentation to read first.