🎓 Is AI killing graduate jobs?

That’s the headline question and we've posed a lot of answers in one direction, but we're a fan of balance so wanted to share what we saw in the FT.

Since ChatGPT’s 2022 debut, job postings for degree-requiring entry-level roles have collapsed by almost two-thirds in the UK. Finance, tech, accounting—once reliable routes into the middle class—are all down 50–75% on pre-pandemic levels. But here’s the catch: it’s not all AI’s fault.

📉 AI’s just the headline act—offshoring, austerity, and economic stagnation are playing the full set

• Offshoring is back. Companies are outsourcing support roles and higher-skill remote tasks.

• Post-Covid over hiring is being “corrected” (translation: hiring freezes and sometimes not so quiet layoffs).

• Economic uncertainty means employers are skittish about committing to training junior staff.

In short, AI might be the poster child, but it’s being used to justify a whole lot of cost-cutting that was coming anyway.

🧠 But AI is changing expectations
Firms are hiring fewer grads, and expecting more of them. You’ll need to onboard faster, take on broader tasks, and yes, be an “AI native” who can work smarter with fewer resources.

Freelance platforms like Fiverr are seeing a Gen Z surge—young people building careers outside of traditional roles, armed with prompts, Canva, and ChatGPT.

⚠️ Now let’s talk about who this hits hardest
The erosion of entry-level roles doesn't fall evenly:

• Underrepresented groups already face barriers to internships and elite networks.

• With fewer formal grad roles and more “figure it out yourself” freelance paths, those without safety nets get left behind.

• And let’s not forget: if you’re Black, Brown, disabled, or working class, you’re more likely to be paid less even when you do land a role.

🚨 So what?
If we don’t invest in accessible career entry points, we’ll recreate the same hierarchies that DEI was meant to disrupt—just in more precarious, gigified form. “Learn to prompt” might be the new “learn to code”, but unless the pipeline is widened, the only thing we’re scaling is inequality.

🎓 Meanwhile, graduates are caught in a Catch-22:
Too junior for senior roles. Too AI-disrupted for junior ones. Too broke to freelance full-time. And too tired to listen to any more podcasts called The Future of Work.

💼 Sovereign AI, Bargain Bin Salary

The UK government has opened applications for a Head of Ventures to help make the UK a “global leader in frontier AI.” The twist? It pays less than £80k. Yes, less than what many entry-level engineers earn at Meta, and a fraction of what top AI talent is making at OpenAI.

Let’s be clear: OpenAI offers base salaries of $200k+ for researchers and engineers—with total compensation packages (including equity) exceeding $800k–$1m. Meta’s new superintelligence team is rumoured to be hiring with packages around $500k–$700k, particularly for those working on LLMs or infrastructure.

So asking someone to design national AI strategy, cut deals with Silicon Valley giants, and run policy for the UK’s "sovereign advantage"—on a mid ranking civil service salary—is more punchline than pipeline according to many in cited in the sifted article

🧠 “Early career” gaslighting
Matt Clifford, who helped design the UK’s AI strategy, says the role is perfect for someone "early in their career." Sure. Because junior staff are totally ready to negotiate compute infrastructure deals with Microsoft and coordinate investment strategies with OpenAI.

📉 DEI, meet DPI (Deeply Poor Incentives)
It’s also a perfect example of how bad pay quietly reinforces exclusion. While public service is often undersold as a “mission over money” play, we know from experience that underrepresented communities are disproportionately underpaid—and often told to be grateful for the opportunity.

Let’s be real: how many Black, Brown, disabled, or working-class technologists can afford to take a job that pays half what the private sector offers—especially when they’re already statistically underpaid? DEI isn’t just about hiring intentions. It’s about building structures that let diverse talent compete and lead. Compensation is a part of that structure.

👀 The big picture
This role is supposed to help the UK define its AI future. But right now, the offer says: build the strategy, make the deals, spin the vision—on a salary half of what you'd expect in the private sector

If the UK is serious about AI sovereignty, it needs to start acting like it. Because right now, the only thing sovereign about this approach is how disconnected it is from global tech realities.

🔗 Read the full story on Sifted

🛑 Age Checks for Porn, 📈 VPNs Up, and 💬 the Real Debate Nobody Wants

As of 25 July, the UK has officially moved past the “click if you’re 18” era of online porn and introduced actual age verification measures under the Online Safety Act. Now, if you want to visit sites like Pornhub, 6000 other sites offering similar content or even things like Reddit, you might need to upload a selfie, share your passport, or get your face scanned by an AI.

Unsurprisingly, the real winner? VPN apps. They’ve rocketed to the top of the UK App Store with downloads up 1,800%, as users scramble to dodge ID checks by pretending to be somewhere else entirely (source).

🤳 What’s Being Asked of Users?

In a push to shield children from explicit material, Ofcom has outlined seven ways websites can verify age — ranging from credit card checks and facial scans to open banking and mobile operator verification. Most of these sound like something you'd expect at a border crossing, not a porn site.

Even tech-savvy young people aren’t impressed. Some are pushing back not on principle, but on practicality — no one wants to hand over biometric data to access five minutes of questionable entertainment.

⚠️ But Here’s the Catch..., Privacy experts are sounding the alarm. Some free VPNs harvest data. Age verification providers aren’t regulated. And in some cases, users are being forced to hand over personal documents to foreign companies. What could possibly go wrong?

As James Baker from Open Rights Group put it: "You're sending a passport scan to access porn. What happens if there's a breach?" LGBTQ+ users and those who aren’t out face even more risk of being outed by metadata leaks — or worse, hacks.

And then there's the cultural whiplash. As James also noted: “In the ‘90s, parents let kids watch violent films — just not the sex scenes.”😅

This policy might reduce how easily kids stumble across adult content. But it’s equally likely to:

• Encourage more underground access — including unregulated and darker content.

• Erode public trust by normalising facial scans and ID uploads.

• Make adults wary of visiting legal, regulated sites, pushing them toward fringe or less secure alternatives.

In short: The government is fighting for child safety with the UX of a digital border checkpoint with a big (walk in this lane if you don't want to), powered by patchy oversight and tech most people don’t understand😂

So yes, protecting kids matters. But if the solution teaches teens how to use a VPN before they’ve sat GCSE History, we might want to ask: are we actually winning?🤨

But what about Banning the VPN's well...? Sure — if your fantasy involves the UK government declaring war on remote workers, journalists, students doing library access from home, cybersecurity professionals, and, oh yes, literally anyone who wants to watch The Office on US Netflix. VPNs are the Swiss Army knives of the internet — used for everything from protecting your privacy to making dodgy hotel Wi-Fi slightly less terrifying. Blocking them because teens might use them to dodge porn filters is like outlawing the car because someone once used one to rob a bank. Not to mention, authoritarian regimes love banning VPNs — so if the UK’s keen to join the digital hall of shame with China and Iran, it’s a stellar start.😳

EE, The mobile network giant has announced new SIM-only plans designed to help parents protect their kids online by blocking adult or harmful content—but only when the phone is using mobile data (4G/5G). 🤔

As soon as the phone connects to Wi-Fi (you know, like at home, school, cafés, or literally anywhere), the filters vanish like your hopes of a quiet dinner.

The idea is that depending on your child’s age, the SIM will apply different levels of restrictions—stricter for pre-teens, looser for older teens—and slow down streaming to reduce screen time. All very noble. But here’s the thing: these filters only work when kids use the internet through EE’s mobile network. As soon as they connect to Wi-Fi, those protections don’t apply—because Wi-Fi is controlled by your home broadband provider, Sometimes not EE.😐

So in practice, this is a bit like locking the front door while leaving every window open and assuming your teenager won't notice. (Spoiler: they will.)🚪

EE is also offering in-store “digital MOTs” for parents—basically a chat to explain how all this works. Which, to be fair, might be needed when the default parental tech strategy is “just Google it, then panic.”

Ultimately, this may be more about making parents feel better than actually stopping kids from seeing what they shouldn’t. After all, the first thing most teens learn about smartphones—after their PIN—is how to find the Wi-Fi password.👶

Sources:
🔗 VPNs soar as porn age checks arrive
🔗 The BBC’s full debate on age verification
🔗 What sites are doing now
🔗 How the age checks work

📉 So what?

This is the UK’s latest attempt to slap a plaster on the gaping wound that is online safety—except this plaster might just leak your passport photo. In trying to protect kids from explicit content, we've opened a whole new can of worms around mass surveillance, dodgy data brokers, and pushing people into unregulated digital back-alleys. It’s a classic case of good intentions, clunky implementation. Because if the outcome is a nation of VPN-savvy teenagers still watching the content and adults uploading their ID to shady third-party firms just to watch consenting adults on the internet… have we really won?

🛑 Narrr you don't say

In the award for pointing out the blindingly obvious, the UK’s Competition and Markets Authority (CMA) just officially declared that Apple and Google have an “effective duopoly” on the UK smartphone market.📱

Groundbreaking stuff. Next up: water is wet, and Liz Truss didn't last as long as the Lettuce.😅

It actually got us thinking, can you actually name another mobile operating system, because we struggled.

But why say that? Well, the CMA’s concern? That the two tech giants use their grip on mobile OSs (iOS and Android, i.e. all the phones) to strong-arm developers and consumers — from app store fees to default browsers to digital wallets. And while they’re not (yet) forcing Apple to allow alternative app stores, the CMA is laying out roadmaps for how both companies should start acting differently.🤨

What was the push back?◀️

• Apple says the proposals are a threat to privacy, security, and — wait for it — their ability to “give away technology to foreign competitors.”

• Google is “disappointed” and wants regulation to be “evidence-based” and not “a roadblock to growth” (read: profits).

• Epic Games’ CEO called the current app store ecosystem “as vibrant as a Soviet supermarket.” Ouch.

But a bit of context◀️

• The UK is late to the game. The EU already forced Apple to open iPhones to competing app stores and well if it's taken them months to do an investigation to conclude Apple and Google have an “effective duopoly” on the UK smartphone market, then maybe we need to look about how tax payer's money is used

• The CMA is one of the few UK regulators still (barely) functioning — unlike Ofwat, which watched s💩t literally pour into rivers.

• Apple and Google made over £9.9bn combined for UK devs in 2022 — so any changes to fees or access would shake the table.

The UK finally clocked what every developer and tech analyst has said for the best part of a decade: you can’t compete with Apple and Google unless you are Apple or Google. Whether the CMA actually does anything bold is another story. For now, we’ll be here, waiting for the day you can delete Safari off your iPhone.📲

📖 Source: The Guardian – CMA to act on Apple and Google mobile platforms

📖 Also covered by: BBC News – Apple and Google have 'effective duopoly'

🔍 So What?

This isn’t just a niche spat between regulators and trillion-dollar tech firms. It’s a critical moment for the future of digital competition in the UK — and whether regulators have the teeth (or political will) to actually do anything about Big Tech dominance.

The Colorintech team got to the conclusion that, they should probably spend their time on something else given how late they are

📤OpenAI + UK Government = A Familiar Consulting Playbook📤

So OpenAI just signed a "non-legally-binding but definitely MOU-heavy" agreement with the UK government to inject AI across public services—from classrooms to courtrooms. It promises productivity, prosperity, and maybe a few quietly automated civil servants.🤖 But if you’ve spent more than 10 minutes around a Big 4 consultant or tech strategist, the playbook is all too familiar:

1. Phase One: “Let us help you figure this out”😅
   OpenAI rolls in like the AI cavalry, offering free advisory support to diagnose “how AI could help.” (Translation: we'll embed ourselves like helpful parasites, map your data pipelines, tech debt, and inefficiencies under the guise of partnership and... discovery.)

2. Phase Two: “Guess what? We have just the solution!”
   After months of embedded insights and a few white papers, lo and behold—OpenAI already has the exact model, the fine-tuning services, the deployment toolchain, and the training modules the civil service needs. And guess what? They’ll help you write the RFP that only they could possibly respond to meaningfully. Because who else understands the problems like the people who designed them?

3. Phase Three: “Now let’s manage and scale it (also, forever)”😅
   After a few successful pilots and lots of productivity graphs, OpenAI is suddenly managing AI infrastructure in Whitehall, popping up in school IT departments, and casually co-writing policy on AI ethics—all while training the next GPT model on public sector data “to better serve the public.”

As Foxglove and others have rightly pointed out: this isn’t just a tech partnership. It’s the soft sell industrial complex in action—where consultation, product, and implementation are all bundled into a sleek, investor-friendly AI-as-a-Service package. Today it’s “increased productivity,” tomorrow it’s a fine-tuned model trained on NHS notes and DWP complaints.💸

📉 So what?

Have we entered the world of not just buying tools; but selling off institutional memory. If government becomes dependent on vendor-trained, vendor-managed AI—especially one with the power of OpenAI—it strikes as outsourcing public services. The UK could wake up five years from now with ChatGPT writing court summaries, diagnosing benefit fraud, and deciding school placements—and no meaningful way to audit, contest, or replace the underlying models without going back to the same vendor, cap in hand.

Oh and what does that mean for diversity? We'll have to ask Open AI about their teams but if we've already seen attempts to remove the public good part comes with a not for profit out of the equation

Link for citation:
OpenAI and UK sign deal to use AI in public services – BBC News

🩸 From “How Often Do You get down?” to $190 Billion in Damages

Just when you thought data privacy cases couldn’t get more personal, Flo—the menstrual and fertility tracking app—has entered the chat. And it’s not alone: Meta is right there with it, defending against a class action in San Francisco over claims the companies secretly shared millions of women’s most intimate health data with third parties.😱

Yes, allegedly: period dates, yeast infections, sex frequency... all piped from your phone to Meta’s servers like a real-time hormonal mood board. And while Meta insists it never “intended” to receive this data, the class action trial is now looming—with up to 38 million users in scope and California’s $5,000-per-violation statute on the line.💰

💸 Do the math:

• 38 million users × $5,000 = $190 billion

• Or, if you count every app entry as a violation? Flo says we’re looking at “quadrillions.” (Yes, with a “Q”)

This isn’t just a GDPR whoopsie. The Flo class action is shaping up to be one of the biggest privacy showdowns of the decade—and maybe, just maybe, it’ll actually make it to trial. Google’s already opted out with a quiet settlement. Meta and Flo? They’re rolling the dice.🤔

🚩 But Wait—You "Consented"

Both companies say users implied consent by accepting the terms of service—somewhere between page 42 and the depths of your disinterest. They also say data was “de-identified.” (Translation: It had a number instead of your name. But your reproductive cycle was still up for grabs.)

Judge James Donato wasn't buying it. He ruled the harm was real: “like stealing a personal diary in 1916”—only now the thief is an SDK baked into your period tracker.👀

📉 So what?

This case is about more than digital eavesdropping on your ovulation. It’s about how casually tech companies treat consent—especially when it comes to women’s data. This is intimate health information from an app built for women, turned into ad targeting fuel allegedly without meaningful permission or oversight. It’s privacy washed in pink branding.

It also exposes a deeper issue in AI and data collection: how much risk is silently borne by women, queer folks, and others who use these tools to manage their health. While Silicon Valley grooms its next AI assistant, real people are wondering if their fertility logs are being used to optimise Digital ads or worse—sold, leaked, or subpoenaed. (Very topical given the Online Safety Act)

This trial won’t fix the ecosystem. But it might finally force the question: What does meaningful consent look like in an age of default surveillance? And how expensive does “oops” need to get before folks start reading their own privacy policies?

🔗 Full story via Reuters

📈 The tools behind the tech📉

📦Product📦

• Mistral launches deep research

📏Design📏 

• A Friendly introduction to SVG

👩🏿‍💻Code👩🏿‍💻

• How anthropic are using Claude code

🏢The business behind the tech🏢

• Tesla bought $16bn worth of chips from Samsung
• China wants to collaborate
• The US action plan

A paid tech secondment

Interested in a paid for secondment to get some great industry folks to support Government (and it’s a great learning experience for the companies too, particularly ones who have work with the government! 👀)

Here is more info: https://www.turing.ac.uk/work-turing/open-source-ai-fellowship-call-2025

Progress in public

Progress in Public: What Is the Civil Service? Your Guide to Digital and Data Careers in Government? 

In case you didn't, it's on the 31st July at 12:00 - 13:30 BST, and it's on Zoom so you can join us from anywhere in the world!  

This webinar will be a great opportunity for you to find out more about Civil Service careers and how to get a job in the public sector. It's particularly relevant for those interested in: 

Engineering, including: DevOps Engineers, Test Engineers, and Specialist Infrastructure Engineers;

Developers, including: Software Developers and Frontend Developers;

IT Professionals, including: Network Architects, Technical Architect, Data Architect

Interaction Designer;

Service Transition Manager.

If you’re interested, please RSVP by signing up on our Luma Page below: 

https://lu.ma/progressinpublicone

😃What we are consuming😃

⚽England's women's won

📺‘South Park’s most furious episode ever'

💔Gwyneth Paltrow appears in bizarre Astronomer video as ‘spokesperson’ after CEO caught in Coldplay kiss cam scandal

🌏The Caribbean islands that give you a passport if you buy a home

🗳️Meta to stop selling political ads in the EU from October

🔍Google's new AI search

📜Decrypting Dynasties