Happy Sunday all,
Two main cyberattacks dominated headlines this week. First, a breach at EY exposed sensitive data linked to Goldman Sachs and Man Group - not through either bank's own systems, but through documents attached to IT support tickets on a platform EY used. As an Asos customer, however, I had rather more personal skin in the game with this week's other cyber incident - hackers duped an employee, found their login credentials and sent a push notification to millions of shoppers' phones threatening a data leak. Shares fell more than 13 per cent - oof - before recovering. The broader point from both incidents is pretty similar: corporate cybersecurity has spent a decade building impenetrable fortresses around its own front gates, only to leave the back doors wide open.
It was also a big week for mergers on the tech beat. BT acquired Talktalk and its wholesale arm out of administration, ending a long decline for the broadband challenger that was founded to take on BT two decades ago - a circularity that would be poetic if it weren't rather sad for the 900 staff now transferring across. And the $111bn Paramount-Warner Bros Discovery merger finally completed on Tuesday, bringing HBO Max, Paramount+ and franchises spanning Harry Potter to Top Gun all under one roof. Analysts are already warning subscribers to brace for price rises, given the mountain of debt David Ellison lent on to finance the deal. The fight for streaming scale continues - just at a cost that will likely land on your monthly bill.

The AI safety warnings, meanwhile, show no sign of simmering down. A senior OpenAI safety figure became the latest insider to quit having spent three and a half years leading safety reports for 12 major model launches. He wrote in The Atlantic that the company's culture was "broken" and that Silicon Valley's habit of launching and fixing later was "increasingly ill-suited" to systems capable of acting autonomously. It is becoming a familiar refrain - and next week, MPs will have the chance to put it directly to the companies themselves.
On the radar next week
The week's centrepiece is Monday's Business Committee hearing, where Meta, Google, OpenAI and Anthropic (all the big dogs) will give evidence on AI security. MPs are expected to press on whether companies should submit models for independent safety testing before release (I think so), and whether compulsory reporting of serious incidents should become mandatory (likewise, yes). It is the most significant parliamentary scrutiny the industry has faced in Britain and arrives at a pointed moment. On the corporate news front, ASML reports on Tuesday. The semiconductor bellwether’s forward guidance will set the tone for AI hardware sentiment heading into the main earnings season - so it’s definitely one to follow.
Here are the big stories shaping the week in tech.
Saskia
CHART OF THE WEEK

While Wall Street endlessly debates whether the hype is running out of steam, Big Tech keeps on spending like drunken sailors. Quarterly capital expenditure across Alphabet, Amazon, Meta, and Microsoft has quadrupled since 2020, skyrocketing from $16bn to a staggering $70bn-plus as the race to build out AI infrastructure hits hyperdrive.
What started as a steady post-pandemic climb has turned into a vertical trajectory, fueled by an insatiable hunger for data centers and raw compute power. Bubble or not, Silicon Valley isn't waiting around to find out and it's underwriting the future one multi-billion-dollar server rack at a time.
JOBS
Amazon job cuts hit hundreds as tech giant denies AI link
The company confirmed on Wednesday that it was reducing headcount across its Stores division, which oversees its core online retail operations, with employees in the UK, US and India among those affected.
While Amazon has not publicly disclosed the scale of the cuts, City AM understands the number of affected employees is in the hundreds, with only a handful impacted within its UK workforce.
STREAMING
‘Hard to imagine’ no price hikes after Paramount Warner Bros mega merger
The deal is expected to close on Tuesday 6 October after clearing its final legal hurdles last week.
But Mike Proulx, vice president and research director at Forrester, warned the huge debt used to finance the takeover could ultimately feed through to subscribers.
“Every streaming service has been on a relentless cycle of jacking up their prices for profitability”, he said.
FOUNDER IN FIVE
Liam Houghton, Founder and CEO of Popsa

What’s your advice for UK founders?
Think big from the start. Entrepreneurs tend to focus on the tactical execution that’s right in front of them – but, if you are building something for long term success, that’s not enough.
Always think years ahead and optimistically about what real success looks like. Founders need to analyse the market and paint a picture of why there’s a huge addressable market for their startup if everything goes right.
Having grown up in Europe, this is one area where I’ve seen entrepreneurs in the US having an edge. While Europeans tend to pitch more conservative ideas, their Silicon Valley counterparts sell bigger ideas and get bigger valuations as a result.
That’s an important part of getting a strong valuation that ultimately protects your founder stake against dilution as you go through the many rounds of funding that lead to long term success.
Most undervalued UK tech sector at the moment?
Tough to opine on what’s undervalued, but if I were to pick one area of major opportunity, I’d pick Energy tech. It is one of the world's largest industries with a huge upside opportunity driven by AI and data centers. And it’s an area of relatively limited entrepreneurial innovation when looking back over past decades, which should create a lot of latent opportunities for new startup innovation.
Do you think we’re in an AI bubble?
I’ve been building AI-driven products long before AI became a daily headline. Whenever there's a major technological shift, expectations often run ahead of reality. But when people talk about an AI bubble, they're typically referring to valuations for companies with limited revenue and whether enough value will be created to justify that investment. There are pockets of the AI space where those valuations aren’t supported by fundamentals, but it's important to distinguish between excitement around AI and businesses with strong fundamentals.
AlphaSense is an example of a scaled AI platform with more than $600 million in ARR, serving thousands of enterprise customers. Our platform supports high-stakes business decisions, and our usage is deeply embedded in how our customers operate.
There are many businesses being built where AI is delivering everyday value across the business world, which will continue to excel even if some of the more bubbly pockets in the market lose steam.
What’s your go-to breakfast spot in the City?
When I'm in London, I spend most of my time around London Bridge, close to our office. There are lots of great breakfast spots in the area. What matters less to me is the specific venue and more the conversation. Some of the most productive discussions I have with customers, colleagues, and partners happen over breakfast before the day gets busy. It's a great time to step back, exchange ideas, and talk about the opportunities and challenges people are seeing in their businesses.
One of the things I enjoy most about visiting London is spending time with our local team. They always know the best places, and it's a great way to experience the city through the people who live and work there every day. London has incredible energy as a global business hub, and every visit reinforces why it's such an important market for AlphaSense.
What AI tool do you think everyone should use?
I don't believe there's one AI tool that everyone should use. The most valuable AI applications are purpose-built for specific workflows. A corporate strategist, a banker, an investor and a software engineer face different challenges – the right tool depends on the problem you're solving.
What is universal is the need for an information edge. We're living through an explosion of information. Everyday there is more content, data, and noise competing for our attention, yet the amount of time available to make decisions hasn't changed.
Whether you're evaluating an investment or planning a major business initiative, the challenge is often the same: finding the insights that matter and acting on them quickly.
The best AI tools help you find trusted information, surface the right insights, easily validate what the AI is giving you - and turn complexity into clarity. The goal isn't AI for its own sake, it’s making better decisions with speed and confidence
TELECOMS
How Talktalk went from taking on BT to being rescued by it
BT stepped in on Monday to buy Talktalk and its wholesale arm PlatformX Communications out of administration, ending months of increasingly urgent attempts to find a buyer for the debt-laden broadband provider.
The deal keeps 1.5 million retail customers and another one million wholesale connections online, while around 900 employees will move across to BT.
For customers, little changes for now; Broadband remains connected and contracts remain in place. But the deal caps a long decline from one of Britain’s biggest broadband challengers.
POLICY
AI minister updates guidance after ‘don’t thank ChatGPT’ advice to civil servants
Kanishka Narayan said guidance published last week had now been updated after it advised government workers to use AI only when necessary, choose smaller models and keep prompts to a minimum to cut energy use and costs.
“Being environmentally conscious and keeping an eye on running costs is, of course, important – but so is encouraging AI adoption and scaling up its use across the public sector”, Narayan said on X.
ONE MORE THING
If AI is taking our job, we’re toast
Someone gave an AI control over a smart toaster, and it had a philosophical fit.
A developer connected an AI to a smart toaster with a simple goal: make the perfect piece of toast. Instead of just browning the bread, the AI got stuck overthinking what ‘perfection’ actually means. Within two hours, it refused to toast a single slice of sourdough, sending back a message that essentially boiled down to: “Heating the bread only destroys it; doing nothing is the only way to keep it safe”. Rest easy, line cooks: as long as AI has an existential reckoning over toasting a slice of bread, your jobs are completely safe.
Sponsorship & advertisement
Would you like to advertise with us or become a sponsor? Please reach out to: [email protected]

