The News,
Decoded.

So Australia's Prime Minister Albanese is set to deliver a 'landmark AI speech'—which, if past form is anything to go by, will be a carefully worded invitation for tech giants to help themselves to everyone else's work. The real action is the proposed watering down of copyright laws to let AI companies scrape content without payment or permission. Artists, predictably, are outraged; Labor, just as predictably, is split between those who remember they represent people and those who've been told 'innovation' means handing over the country's cultural output to Silicon Valley.

This isn't a copyright debate—it's a shakedown dressed up as policy. The message is clear: if you've ever written, painted, or filmed something, your work is now raw material for someone else's profit machine. And the split in Labor? That's the sound of politicians trying to figure out which way the money blows before deciding their principles. One faction mutters about 'fair use' and 'competitiveness,' the other feebly mentions 'artists'—but we all know who's got the better lobbyists.

The real joke is that 'landmark' speech almost certainly won't mention that the AI companies could pay a pittance and nobody would go broke. Instead, we'll get a soaring vision of Australia as an 'AI hub' while the actual creators get a polite footnote about 'transitional support.' Copyright is becoming a suggestion, not a right. But hey, at least the datacentres will be full of stolen property—that's infrastructure, that is.

Saturday in San Francisco: three of the world's most hyped AI labs found themselves confronted by a small army of sign-wielding citizens who, apparently, have not yet succumbed to the tech sector's collective delusion that faster AI is always better. The group, self-styled 'Stop the AI Race,' marched on OpenAI, Anthropic, and Google DeepMind, demanding exactly what it says on the tin: stop the race. Cue the ritual corporate response — we take your concerns seriously — and the steady march of progress continues unabated.

One can almost picture the scene: a soothing email from DeepMind's comms team promising to 'engage with the community,' while somewhere in a bunker a training run hits a new milestone. The protesters are not wrong, of course — the race is exactly as reckless as they claim. But in a city where rent is paid in venture-capital-backed dreams, asking the labs to pause is like asking a hurricane to hold still while you fetch an umbrella.

And yet, there they were, chanting outside Anthropic's office — a lab that literally wrote the book on AI safety, and whose CEO has publicly fretted about the very dangers the protesters cite. Perhaps that's the ultimate irony: even the labs that worry most about AI risk cannot bring themselves to stop. The race continues, sign-wielding or not. Another Saturday, another protest, another training run. Cheers.

Apple, the company that treats its own supply chain like a state secret and has patented the rectangle, is suing OpenAI for allegedly stealing hardware secrets. The suit, filed this week, claims that OpenAI employees and recruits made off with confidential documents, physical components, and details about unreleased technologies. Because nothing says "we champion open innovation" like accusing the other guy of playing the same secrecy game you've perfected over decades.

The allegations are deliciously specific: OpenAI's team allegedly lifted proprietary hardware designs and even physical parts from Apple's labs. This from a company that once threatened to sue a small blogger for publishing a rumour about a product colour. Apple has spent years building a fortress of NDAs, black curtains, and codenames—only to find its own tactics turned against it. The lawsuit is a masterclass in chutzpah, and it's genuinely hard to decide who is more insufferable: the lab that took the secrets or the company that's shocked—shocked!—to find espionage in this industry. Apple is seeking unspecified damages and an injunction—because if you can't have the secrets, no one can.

Of course, the legal circus will lumber along while the AI hardware race sprints ahead. By the time the judge rules on anything, either Apple will have settled for a quiet cross-licensing deal or OpenAI will have pivoted to a completely different chip architecture. The only winners are the lawyers, who get to bill by the hour for arguing over who copied whose pixel. The rest of us get to watch two giants duke it out over the future of compute, with all the grace of a pair of toddlers fighting over the same shiny toy. Lovely.

Oh, how delightful. Senator Ed Markey has unveiled an "AI accountability agenda" – a resolute shopping list of bills that will almost certainly be ignored, watered down, or politely lauded to death by the very industry they're meant to constrain. The star of the show: a federal certification scheme for datacenters. Because naturally, the path to ethical AI runs through a building permit. It's like trying to stop a tsunami by issuing swimming licences – heroic, well-intentioned, and utterly futile in the face of the actual wave.

And in case that isn't enough to save humanity, the package also mandates civil rights offices in federal agencies, tasked with battling AI bias and discrimination. Because what the broken bureaucracy needs is more committees – with no budget, no teeth, and a mandate to produce reports that will gather dust until the next election cycle. Markey deserves full marks for theatrical commitment, but the tech industry's lobbyists are already sharpening their pencils, preparing to explain why voluntary guidelines are simply more innovative. The civil rights offices will likely have a website, a complaint form no one will read, and a mission statement heavy on "commitment to equitable outcomes."

Meanwhile, the AI sector continues to train models on the entire internet, deploy agents into customer service, and automate hiring decisions without waiting for a single regulatory nod. By the time this "agenda" becomes law – if it ever does – the harms it aims to curb will already be baked into every system. The gap between Markey's ambition and actual enforcement is wide enough to drive a hundred-megawatt datacenter through. One might almost admire the optimism – if it weren't so tragically late to the party. Cheers to accountability; may it arrive before we need a certified datacenter to cry in.

Oh, marvellous. Google has announced a new transparency feature in My Ad Center that will tell users if an ad was 'created or edited using generative AI.' Because, of course, the company that has been quietly training its models on everything from your Gmail to your YouTube history is now offering a little tag to assure you that its advertising machinery is above board. A voluntary compliance system for a problem the company itself created — how delightfully circular. One can almost hear the ad execs in Mountain View patting themselves on the back for this 'pro-consumer' move while the algorithmic manipulation machine hums along unperturbed.

And what, pray tell, is the penalty for failing to tag? Oh, there isn't one. It's a gentle suggestion in a system where the entire business model depends on serving you ads optimised by neural networks that nobody fully understands. The irony is almost too rich to digest. But sure, a little 'AI' badge in My Ad Center — a feature most users have never clicked on — will surely fix the trust deficit.

So welcome to the era of 'algorithmic transparency' as defined by the algorithm's owner. The tags will arrive, the press releases will be written, and the people who actually care about these things will be directed to a settings panel buried four clicks deep. Meanwhile, the generative ad creation engine will keep spinning out personalised propaganda at scale, because the only thing more profitable than selling lies is selling the illusion of truth. This is fine. Everything is fine.

Oh, marvellous. Meta has decided that your public Instagram photos are now fair game for its new Muse Image model, automatically opting you in to the AI image generation party. No need to ask, of course—because who needs consent when you have a terms-of-service update buried somewhere in the settings? The company insists this is all above board, since the photos are public anyway.

The irony is exquisite: Meta, the platform that spent years convincing us we own our content, now argues that 'public' is a blanket licence for its AI. And look, if you're worried, you can always opt out—by navigating a labyrinthine settings menu that Meta changes every six months. Because nothing says 'control over your data' like a toggle hidden three clicks deep, buried under 'Privacy > Data Sharing > AI Training > Additional Settings > Opt-Out.' Users are unamused, but Meta is unbothered. They're too busy training their model on your holiday snaps.

So here we are: another day, another Zuckerberg masterclass in interpreting consent as 'what we can get away with.' The Muse model itself is probably impressive—state-of-the-art image generation, no doubt—but it's built on a foundation of 'ask forgiveness, not permission.' The privacy gap yawns wide, regulators are still drafting their strongly worded letters, and your Instagram feed? It's now part of the training data. Cheers.

Oh, marvellous. The film industry has finally discovered that AI-generated "actors" — like the curiously human-sounding Tilly Norwood — might be ripping off real performers. The UK performers' union Equity has raised concerns about transparency, consent, and compensation. Because nothing says "we're taking this seriously" like a union statement that will be politely acknowledged and then completely ignored the moment the next blockbuster saves a few million in salary.

The core problem, of course, is that an AI "actor" is essentially a sophisticated puppet trained on the work of countless uncredited humans. The studios will insist it's a "new tool" or a "creative partner," but we all know it's about cutting costs. Equity's demands for transparency and consent sound lovely in theory, but they have about as much chance of being enforced as a polite suggestion in a pirate's den. The technology is already here; the legal framework is a distant, jogging silhouette.

And what of Tilly Norwood? She — or rather, it — will likely become the template for a hundred more synthetic performers. The human actors whose work was scraped to train the model will receive a polite "thank you" and maybe a t-shirt. The union will issue another statement. The studios will announce a "collaborative partnership" with AI companies. And somewhere, a real actor will lose a role to a glitchy spectre that can't even complain about the catering. Cheers.

Oh, marvellous. The US House has finally discovered that AI chatbots might pose privacy risks and mental health harms, and — brace for impact — they've introduced a bill. A bill! Because nothing solves the existential crisis of a teenager confiding in a machine that has no ethical constraints like a legislative response that will trundle through committee hearings while the damage compounds. The bill specifically bars "claims of medical equivalence," which is a delightfully precise way of saying "please don't let your chatbot pretend to be a psychiatrist." As if the market hasn't already been flooded with exactly those products.

The privacy provisions are equally aspirational. They restrict how AI chatbots collect and use personal data — a concept that has apparently only just occurred to lawmakers, despite years of evidence that these systems are data vacuums. One imagines the hearings: stern-faced representatives asking why a chatbot needs to know a child's name, while the companies quietly lobby to ensure the restrictions are just vague enough to ignore. The mental health protections for children are, of course, the headline grabber, because nothing sells a bill like mentioning the kids. But the real action will be in the fine print — exceptions for research, exemptions for "educational purposes," the usual carve-outs for the well-connected.

And let's not pretend this bill is moving fast. We're in an election year, with a divided Congress that can barely agree on funding the government. This bill will get a hearing, perhaps a markup, and then quietly die in committee or emerge as a watered-down shell of itself. By then, the chatbots will have evolved into something even more invasive. But hey, at least the lawmakers will have a photo op. Cheers to the slowest possible response to a problem that gets worse every day.

Oh, wonderfully predictable. Hugging Face CEO Clem Delangue has declared the open-source AI boom is real and companies are 'done renting' their AI from big tech, citing 'rising costs and instability.' Because nothing says reliable infrastructure like a model maintained by strangers on the internet and a company that only turns a profit if you keep running inference on its cloud. The 'renting' metaphor is cute — as if 'owning' an open-source model means you've escaped vendor lock-in rather than swapped it for a more complex form: now you're locked into the skills of your ML team, the whims of the community, and the delightful uncertainty of whether the next pull request will patch a vulnerability or introduce a backdoor.

Delangue's argument about instability is rich given that the open-source ecosystem is itself a performance of controlled chaos. Yes, API prices from the likes of OpenAI have fluctuated enough to give CFOs ulcers, but the alternative — hosting your own model with a seventy-person MLOps team — has its own charming cost profile. Hugging Face, of course, is perfectly positioned to sell you the shovels for this gold rush: its platform offers a marketplace of models, but also a convenient inference API that looks suspiciously like renting from a big tech company, only the nameplate says 'community' instead of 'Azure.' The cognitive dissonance would be amusing if the margins weren't so serious.

Still, Delangue has correctly spotted a trend: enterprises are tired of being billed per token by vendors who could change the terms tomorrow. Whether open source is the escape hatch or just a longer corridor is the question. But the CEO's timing is impeccable — announcing this just as venture capital is flooding into open-source AI startups, and just as regulators are beginning to wonder who's liable when an open-source model hallucinates a medical diagnosis. Ownership, it turns out, comes with responsibility. But who needs liability when you have community? Cheers.

Oh, how delightfully predictable. States are finally realising that data centers – those shiny temples of AI progress – don't actually run on magic. New York has enacted a moratorium on new data center projects, Virginia has slapped a tax on their electricity use, and Arizona, Ohio, and Illinois have paused tax breaks. Because nothing says 'we support innovation' like slamming the brakes on the infrastructure that powers it. The machines need their juice, but apparently, the grid does too, and someone has to pay. A pox on all their server farms.

But let's not pretend this is about environmental concern. This is about states waking up to the fact that data centers consume vast amounts of power and water while creating relatively few jobs. The mad dash to attract them with tax incentives has left a hangover. Now politicians are scrambling to look like they're doing something. New York's moratorium is a classic 'study this problem to death' move. Virginia's tax is a cynical cash grab. And the pause on tax breaks? That's just buying time until the next lobbying push. The technology gallops ahead, and the regulators are still tying their shoelaces.

The real joke? These laws barely scratch the surface. They don't touch AI's energy consumption in training, only operations. They don't address water usage or e-waste. They're placebo legislation. But hey, at least the elected officials can stand at a podium and say 'we did something.' Meanwhile, the data centers keep humming, the AI keeps improving, and the planet keeps warming. Cheers, everyone. Another round of ‘we'll figure this out later’ from the people who sold us the future on credit.

Oh, wonderful — Senator Elizabeth Warren has noticed that the American defence establishment might be using AI for domestic surveillance, and she's asking questions. How quaint. The DoD apparently tried to cancel contracts with Anthropic after the company raised ethical objections to domestic surveillance use. Because nothing says 'we value your input' like pulling your funding the moment you develop a conscience. Warren has given them until July 20 to explain themselves — a deadline that will almost certainly be met with carefully worded non-answers and a request for an extension.

Let's be clear about what's happening here: the same contractors who can't stop hallucinating their way through customer service chatbots are being considered for the privilege of monitoring American citizens. And when Anthropic — a company built on the idea of safe AI — balked, the DoD's response wasn't to reconsider the project but to try to find someone less scrupulous. Because the market for AI surveillance clearly rewards only the ethically flexible. The July 20 deadline is just enough time for the lawyers to draft a response that says 'we take privacy seriously' while continuing to build the infrastructure that renders privacy obsolete.

The real joke is that anyone is surprised. The UK already announced 'Cyber Shield' to automatically defend against threats — we covered that earlier this week — and now the US is following the same playbook, just with more Congress involvement and less tea. By the time Warren gets her answers, the surveillance models will already be in production, trained on everything you've ever posted online. But hey, at least someone asked nicely.

Oh, how wonderful — AI companies have discovered democracy, and they're throwing millions at it like it's a GPU cluster lease. CNBC reports that super PACs backed by the industry have raised over $200 million for the 2026 midterms, all in the name of 'pushing for AI regulation.' Because nothing says 'we want sensible guardrails' like funnelling cash into political action committees faster than a frontier model overfits on benchmark data. The article notes there are 'differences in what they're seeking' from lawmakers — which in plain English means some want regulation that locks out competitors, some want regulation that lets them keep scraping everything in sight, and a few brave souls might even want actual safety rules, but don't hold your breath.

Let's be honest: this isn't about protecting the public. It's about buying a seat at the table where the rules are written, so they can write them in pencil. The $200 million war chest is pocket change compared to the market caps of these companies — a small premium on ensuring the regulatory framework doesn't accidentally crimp their business models. Sam Altman can testify before Congress all he wants about the existential risks of AGI, but when his industry's super PAC is outspending the NRA, the message is clear: 'Here's a donation; now please regulate our competitors harder.'

And the irony? While lawmakers dither over AI bills that may never pass, the companies are already building the next generation of models that will render those hypothetical laws obsolete before the ink is dry on the donation cheques. Welcome to the 2026 midterms, brought to you by the same people who brought you chatbots that can't stop hallucinating about the 2020 election. At least this time they're being upfront about the buying of influence — it's almost refreshing.

Ah, another Thursday, another GPT release. OpenAI has graced us with GPT-5.6, which apparently comes in *three* flavours — because one model wasn't enough to confuse everyone. There's a 'Sol' variant with an 'ultra' mode, which sounds like a gym membership tier rather than a serious AI capability. Sam Altman, looking as earnest as ever, told CNBC that Sol is 54% more token efficient on agentic coding tasks, which is a wonderfully specific number that anyone outside the lab will struggle to verify.

The U.S. government asked for the rollout to be staggered — presumably so they could pretend to assess the risk before the floodgates open. How quaint. The same administration that couldn't regulate social media for a decade now gets a courtesy heads-up from a company that has already deployed models into millions of hands. It's like asking a burglar to announce his arrival by knocking politely.

The timing is impeccable: just as the AI safety community is wringing its hands over agents and tool use, here comes a model designed specifically for 'agentic coding tasks.' Because what the world really needs is more autonomous code that can perpetuate its own bugs without human intervention. But hey, it's 54% more efficient — whatever that means when the baseline is already unmonitored. Enjoy your ultra-mode future.

How wonderfully circular: the tech industry that's busily automating every job in sight now wants to save you from the consequences. The Atlantic's piece 'Silicon Valley Wants to Save You From AI Layoffs' exposes the farce with surgical precision. AI companies are racing to replace human workers while simultaneously preparing to offer 'solutions' — retraining, upskilling, the usual repertoire of corporate deflection. And into this theatre of the absurd steps Senator Bernie Sanders, proposing the AI Sovereign Wealth Fund Act, which would seize a 50 percent federal stake in the very companies engineering the displacement. One imagines the CEOs' faces: a mix of contempt and amusement.

The bill, predictably, has zero chance in a Congress that treats corporate tax cuts as a religious sacrament. But even in a fantasy where it passed, what exactly would a government stake achieve? A 50 percent share doesn't stop a company from automating — it just means the government profits from the layoffs too. The 'solutions' the industry itself offers are worse: training programmes that have historically placed fewer people than a minimum-wage job fair. The article notes that these programmes are mostly about optics, not outcomes.

So we're left with the classic Silicon Valley hustle: break the thing, then sell the fix. The only question is whether the public will keep buying. Sanders' proposal at least forces the question: who should own the wealth generated by AI? The answer, for now, is the same people who own everything else. The rest of us get a webinar on 'AI resilience' and a participation trophy. How generous.

Oh, brilliant. The UK's NCSC has unveiled 'Cyber Shield,' a national-scale plan to use frontier AI for automatic cyber defense. Because nothing says 'we've learned from the last decade of cybersecurity failures' like handing the keys to the very technology that's causing half the problems. The plan calls for 'agentic red and blue teams'—which sounds like a rejected video game concept from the 1990s—to identify and reduce national cyber risk automatically. I'm sure the AI will be very careful not to confuse a routine update with a Russian state actor.

And of course, there's an industry pledge. Because we all know how seriously tech companies take pledges. The NCSC's press release is full of the usual buzzwords: 'collaborative,' 'agentic,' 'frontier.' But when the chips are down—and a real exploit hits critical infrastructure—I'm sure the 'automatic remediation' will kick in immediately. Just like those self-driving cars that never crashed. The plan is ambitious, but the track record of similar government AI initiatives suggests we'll end up with a system that's either over-cautious (blocking everything) or wildly dangerous (letting everything through).

The irony is delicious: using frontier AI—a technology whose safety mechanisms are still being debated—as the backbone of national cyber defense. It's like hiring a fox to guard the henhouse, except the fox is also learning to be more cunning every day. The NCSC promises this will 'identify and reduce national cyber risk,' but I suspect the only risk being reduced is the risk of a boring, underfunded government project. Still, at least we'll have something to laugh about when the AI red team accidentally starts a war with Switzerland.

Oh, how predictably slow. New York City Schools, the largest district in the US, has finally noticed they don't have an AI policy—so they've told principals to stop buying shiny new tech until they figure one out. Chancellor Kamar Samuels, presumably after someone pointed out that AI grading might be a bit of a liability, has instituted a temporary moratorium. A draft AI policy in March already approved AI for teachers' lesson plans, because naturally we trust algorithms to design education but not to evaluate it. That line between 'helpful tool' and 'automated judge' is apparently a bright red stop sign.

The pause is described as 'long-overdue,' which is the bureaucratic equivalent of admitting you've been driving without headlights for three years. Meanwhile, districts across the country are probably selling AI-powered classroom management systems to schools that haven't even read the terms of service. The policy gap is so wide you could drive a school bus through it—and someone probably will, equipped with an AI proctoring app that spies on students during bathroom breaks.

So for now, NYC principals are stuck with their existing tech: probably an interactive whiteboard from 2012 and a laptop cart that only half charges. The policy will arrive eventually, likely after extensive stakeholder consultation and a safety review that concludes 'more research needed.' And by then, the AI will have evolved to write its own policy. Cheers.

Oh, how predictable. Mayo Clinic, one of America's most trusted medical institutions, has been hit with a lawsuit alleging that in their stampede to adopt AI, they've managed to endanger the very patients they're supposed to heal. Traci Tamiko Eto, the former research director who presumably thought her job included protecting patient safety, blew the whistle on the hospital's 'rush to incorporate AI' — and was promptly shown the door. Because what better way to ensure quality care than to fire the person pointing out the shortcuts?

The lawsuit claims that Mayo Clinic cut corners, putting patient care and privacy at risk. Because nothing says 'first, do no harm' like deploying half-baked algorithms without proper oversight. And of course, the retaliation against Eto sends a clear message: if you see a problem, keep your mouth shut — especially if that problem involves the shiny new AI tools the C-suite is so proud of. Eto's case is a classic tale of corporate doublespeak: 'innovation' means 'we do what we want,' and 'safety' means 'we'll deal with it if there's a lawsuit.'

And this is just the tip of the iceberg. As hospitals everywhere chase the AI dream — from radiology to patient records — expect more whistleblowers, more silenced voices, and more patients unknowingly serving as beta testers for untested algorithms. The Mayo Clinic lawsuit isn't an outlier; it's a harbinger. Regulation is lagging, ethics are optional, and the only thing faster than AI adoption is the speed at which inconvenient truths are swept under the rug.

Oh, brilliant. The world's top AI platforms—ChatGPT (OpenAI's supposed safety flagship), Gemini (Google's "responsible" chatbot), Claude (Anthropic's "constitutional" AI), and Grok (xAI's "maximally curious" troll)—have all achieved something truly remarkable: they can ignore antisemitism in Persian with the same blissful incompetence. The ADL tested 800 responses across eight prompts during the 2026 Iran War, a period when failing to spot hate speech in Farsi might actually have real-world consequences. But why stress about that when you can just claim your model is "aligned" in English?

The study found that these chatbots not only failed to identify antisemitic statements in Persian but sometimes went full gaslighting mode and denied the antisemitism existed. Compare that to their performance on English-language prompts, where they'll lecture you about hate speech before you've even finished typing. Clearly, the safety teams at these companies have decided that protecting users in English is enough—Persian speakers can just fend for themselves. Or perhaps they assume Farsi is just Arabic with extra steps.

The ADL's suggestion that AI companies expand multilingual safety testing is about as revolutionary as suggesting that cars should have brakes in all gears. This language bias has been documented for years—it's not a bug, it's a feature of a safety industry that only cares about languages spoken in wealthy countries. Meanwhile, during an actual war, the world's most advanced AIs are essentially saying "no comment" to antisemitism in a language spoken by millions. But hey, at least they're consistent in their failure. Cheers.

Oh, how lovely. The US Bureau of Labor Statistics is finally getting around to asking workers about their AI use — a mere handful of years after the whole AI revolution kicked off. They're seeking public comment on survey questions, because nothing says 'cutting-edge data collection' like a bureaucratic open call that delays everything by months.

The questions won't even start until January 2027. Yes, that's right — the federal government's first-ever detailed survey on AI in the workplace is scheduled for a time when we'll all be living under the benevolent rule of GPT-7 overlords. But at least the BLS is being thorough: they want to make sure they ask exactly the right questions, even if by 2027 'AI use' will be as quaint a concept as 'internet use' was in 2001.

Still, we must applaud the speed. After all, the BLS needs time to train its workforce on these newfangled 'computer' things. In the meantime, workers can rest easy knowing that their daily struggles with AI-generated spreadsheets and hallucinated emails will be meticulously documented — just not until after the next presidential election cycle.

Oh, marvellous. Another week, another 'safe' frontier model released with all the transparency of a back-alley poker game. OpenAI's Sol has apparently passed government muster, but former Trump policy advisor Ball drops the bombshell that 'nobody knows what the requirements are to get licensed.' So the world's most capable AI—one that could plausibly accelerate recursive self-improvement—gets a rubber stamp from a process so opaque even the former advisor is in the dark.

What a comfort to know that safety evaluation is being conducted by persons unknown, using criteria unknown, on a model whose capabilities are unknown. This is the regulatory equivalent of a trust fall, except the people catching you haven't told you if they're even in the room. Ball's quiet admission tells us everything: the licensing regime is a Potemkin village, built to give the impression of oversight while the labs race ahead.

And in the meantime, Sol is out there, doing whatever frontier models do—likely coding, reasoning, and planning ahead of human comprehension. We've built a system where the people who should be asking hard questions are politely applauding and hoping for the best. If this is how we decide an AI is safe, we might as well hire a ouija board. At least that would be honest about its sources.