Britain

Britain woos Anthropic to expand after clash with Pentagon

Britain woos Anthropic to expand after clash with Pentagon

Here is where things stand. The US Defense Department designated Anthropic a national-security supply-chain risk after the company refused to allow its Claude models to be used for military surveillance and autonomous weapons.

A federal judge blocked the designation, ruling it likely violated constitutional protections. The Trump administration is now appealing that ruling. The President, separately, called Anthropic’s leadership “leftwing nut jobs” for holding that line.

Into that opening, Britain moved quickly.

The UK government is courting Anthropic with proposals that include expanding its London office footprint and pursuing a dual listing on the London Stock Exchange. Officials at the Department for Science, Innovation and Technology have drafted the proposals for Anthropic CEO Dario Amodei, who visits Britain in late May on a European customer and policy tour. Downing Street is backing the effort.  London Mayor Sadiq Khan followed up in writing, pitching the capital as a “steadfast” base for the company. The FT broke the story on Sunday.

The proposal on the table is part expansion offer, part diplomatic signal. Britain wants Anthropic in London. It also wants to be seen wanting Anthropic in London, which is a different thing and equally intentional.

The honest subtext, acknowledged privately by officials, is that Britain has no homegrown frontier lab to rival the Americans. The strategy is partnership, not competition. The goal is to tie the best US labs to UK infrastructure, research base, and talent pipeline before other European capitals do.  OpenAI has already committed to making London its largest research hub outside the US. Google is completing a roughly £1 billion King’s Cross campus. The Anthropic pitch fits a pattern.

But this story is not really about office space or stock listings. Those are instruments. The story is about what a government does when a private company refuses a government’s demand and gets punished for it, and another government decides that refusal is an asset worth recruiting.

Anthropic drew a line. It said Claude will not be used for surveillance. It said Claude will not be used for autonomous weapons. The Pentagon designated it a risk for saying so. That sequence is the thing worth sitting with, because it describes something new about where AI sits in the world right now.

For most of computing history, technology was neutral in the geopolitical sense. Governments bought it, used it, regulated it, but the tools themselves did not have positions. What is happening now is different. The major AI labs are being asked to take sides, not rhetorically, but operationally. Will your model help target people? Will it automate lethal decisions? The answer to those questions is becoming a foreign policy matter.

Britain is not offering Anthropic a home because it agrees with every position Anthropic holds. It is offering a home because a company willing to refuse the US military on ethical grounds is a company that other governments can negotiate with. That is valuable in a world where AI is becoming as strategically significant as energy or communications infrastructure.

A dual listing remains, in the words of one insider, “the dream” rather than a realistic near-term scenario, particularly with Anthropic expected to IPO in the US as early as this year. The legal cloud from the Pentagon appeal is still in place, and formal commitments are unlikely before that resolves.

What is not in question is the direction of travel. The AI labs are no longer just technology companies navigating markets. They are entities with enough independent weight that governments court them, punish them, and position themselves around them the way they once did around oil companies or defense contractors.

The question of whether that power comes with accountability, and to whom, and under which legal framework, is one nobody has answered yet. Britain is not answering it either. It is just making sure it has a seat at the table when someone does.

That is what this visit in late May is really about.

Google

Google Launches its Most Versatile Models to Date: the Gemma 4

Google Launches its Most Versatile Models to Date: the Gemma 4

If Google is giving away the same AI that OpenAI charges for, does a $20 monthly subscription even make sense anymore?

Google just dropped Gemma 4, and it feels like a direct hit to the subscription model. For the last few years, the best AI lived behind a paywall. If you wanted the good stuff, you had to pay OpenAI or Anthropic every month.

Google is now giving away a model that runs on your own hardware for free. It is a smart move to turn high-end AI into a basic utility that anyone can use.

The license is the real story here.

Google is allowing anyone to use the code without requiring permission by leveraging the Apache 2.0 standard framework. You can take this model, put it on a private server, and use it to handle sensitive data such as medical records or bank statements.

You never have to send a single byte of data to a third-party cloud. It solves the privacy challenge that has been bothering prominent industries for years.

Gemma 4 is surprisingly versatile.

It handles audio, vision, and text all at once. Because it runs locally, it works in airplane mode. You could be in a remote area and use your phone to translate a conversation or identify a plant through your camera. It removes the lag and the cost of the cloud.

Google’s strategy is simple.

If they can’t be the biggest paid service, they will be the best free foundation. They want every developer on the planet building on their tech. By making the “brain” a commodity, they are forcing competitors to justify their high prices. It is a race to the bottom, and for once, the users are winning.

The elite AI paywall just hit a wall.

Claude

Claude Code Leak on X Directs Scrutiny Towards Anthropic

Claude Code Leak on X Directs Scrutiny Towards Anthropic

Users caught a look into how Claude really thinks- and it thinks a lot about ASCII capybaras and memory pruning.

AI development is transforming industries- and at the very core of where it stems from, it’s changing the coding landscape too. It’s more of a psychological take than a technical one.

We assume that software developers need minimal distraction and high-efficiency tools to code. But, agentic development changed that- it’s the rise of the buddy system in engineering circles. Humans write the code, and the “buddy” helps them through soft errors that AI workflows often instill.

When Claude’s 512000-line code repository was leaked on X, it also revealed a secret April Fools’ gamification feature Anthropic was planning. That was a bigger discovery- a “/buddy” repository.

It drew as much focus as the code itself- how Claude handles shell execution and permissions. Security researchers now don’t have to guess how to break out of the agentic sandbox. And given all the tip-toeing around AI, security through obscurity can’t be the only tactic known.

Now that the agent’s logic is known? It sounds impossible to pull back the harness. However, Anthropic is attempting its best, playing Whack-a-Mole. So far, it has removed over 8k forks from GitHub. However, the consequence of a simple human error is present on several decentralized platforms.

Users can already notice numerous clean-room implementations uploaded on Rust and Python.

This scenario has set history for AI IP: Claude Code has given its competitors a blueprint, even when there was no user data leak. While some will receive access to the downloadable leaked Claude mirrors, others will end up with malware-laden cracked versions.

The black box era of AI just took a huge hit. Now that we have had our glimpse behind the curtains, can one declare with confidence- “we now know how the world’s most advanced AI agent thinks?” Or could there be more to what meets the eye?

Video Apps

Video Apps are Reeling in More Active Users Than Social Media Apps

Video Apps are Reeling in More Active Users Than Social Media Apps

Is it the era of the great digital retreat? Ofcom’s latest report concludes so.

Ofcom recently conducted a survey across UK social media users and noticed a strange dip. Only 49% of the UK adults actually post, share or even comment on social media. That’s a 61% decline since 2024.

The reason, one can assume, is the panopticon effect. Social media was always thought be a space for the ‘now’- the ‘present.’ But if one hasn’t noticed, it’s now used as a record of our past and present followings.

You can also call this: archive anxiety. It’s the fear that a single version of ‘you’ will be used against you today. The migration to more ephemeral media isn’t sudden. People would rather opt for posting on Instagram stories than make grid posts.

But another reason- adults are migrating towards more-video oriented content, such as Reels and TikTok.

Empty consumption over active creation- users merely wish to be passive on-lookers. And scrolling video (with infinite scroll loop) feeds into a dopamine hit that creating the content doesn’t fill.

Video-centric feeds have turned social media into more of cinema- one that’s observed not actively interacted with. But what use is an interaction that’s majorly AI-driven? The study also asserts that over 54% of UK adults leverage AI for conversation.

Because it’s the path of least resistance. The friendship is low-stakes and the AI is more of a one-sided consultant than a friend that requires mutual effort.

Bottom line? Humans are retreating from the chaotic complexities of ‘human’ social media- inching more and more towards a controlled comfort of AI dialogue. And honestly, more than half of them truly hold the belief that social media isn’t good for their mental health.

Their retreat is backed by some compelling arguments. 89% still use the platforms even though they believe its harmful- that’s the hold social media has built on users. That’s the psychological trap that big tech is counting on, to keep users hooked to these echo-chambers.  

Yahoo

Yahoo Repositions its DSP to Attract Downmarket Advertisers

Yahoo Repositions its DSP to Attract Downmarket Advertisers

Yahoo is opening its premium DSP to smaller advertisers, but can an AI-powered blueprint really convince mid-market brands to ditch Google and Meta?

Yahoo is currently rewriting its playbook.

For years, the Yahoo DSP was the playground of Fortune 500 brands with massive budgets and dedicated teams. But as of early 2026, the company is pivoting to attract downmarket advertisers- the mid-sized brands and local agencies that have long felt priced out of premium programmatic tools.

That’s a calculated move to become the primary alternative for those tired of the Google-Meta duopoly.

The shift is anchored by the launch of Yahoo Blueprint, an AI engine designed to do the heavy lifting that requires a fleet of data scientists. The goal is to make the platform self-service and intuitive. Instead of wrestling with complex bid strategies, a small marketing team can now use agentic AI to automate campaign optimization.

Yahoo is betting that by lowering the barrier to entry, it can capture the massive wave of ad spend currently flowing into simpler, but less powerful, social media platforms.

The real draw, however, isn’t just the ease of use; it’s the data.

Yahoo is opening up its “superpower,” i.e., first-party data from Finance, Sports, and Mail, to these smaller players. That gives a regional car dealership or a boutique e-commerce brand the same level of targeting precision as a national retailer.

By integrating its identity solutions directly into a more affordable tier, Yahoo is offering a walled garden experience on the open web.

Of course, this repositioning is also a defensive necessity.

With the recent merger of LINE and Yahoo Japan’s ad platforms, the global entity is searching for scale. They need more than just a few whales- a school of mid-sized fish to maintain the ecosystem.

Yahoo is betting that by becoming the approachable elite platform, it can finally turn its legacy data into a modern gold mine.

OpenAI

OpenAI Closes $122bn Funding Round, Achieves $852bn Valuation

OpenAI Closes $122bn Funding Round, Achieves $852bn Valuation

OpenAI just raised a staggering $122 billion. If it’s redesigning its roadmap- why does it smell of desperation to stay ahead of the hype cycle?

OpenAI just closed a massive funding round, which could be a small country’s GDP.

It isn’t just about paying the electric bill for servers. It is a loud, expensive bet that the path to AGI is paved with sheer, brute-force capital. By bringing in a mix of sovereign wealth funds and tech giants, Sam Altman is effectively trying to out-spend the laws of diminishing returns.

The real story here isn’t the number of zeros, but the shift in OpenAI’s identity.

The company is reportedly moving away from its complex non-profit roots to become a fully for-profit entity. This change is the price of admission for such a massive check.

Investors are no longer satisfied with saving humanity as a mission statement; they want a clear, legal path to a return on their hundred-billion-dollar investment. The road to AGI requires a level of commercialization that the founders once fought against.

However, there’s a lingering tension behind the hype.

As OpenAI scales, its hunger for data and power is hitting physical limits. We are seeing a pivot toward synthetic data and custom nuclear power deals because the internet is simply running out of fresh human thoughts to feed the machine.

This funding round buys them time to solve those engineering hurdles, and it places a massive target on their back for regulators who worry about a single company holding the keys to the future.

We are entering a phase where the AI race is determined by who has the deepest pockets. The $122 billion is a signal to rivals like Google and Anthropic that OpenAI is willing to burn through cash merely to become the pack leader.

The question is whether all that money can actually buy the breakthrough they’ve promised.