🏥AI gender bias in health🏥

A study from the London School of Economics (LSE) has found that AI tools used to summarize long-term care records are showing clear gender bias. And while AI companies claim to be leading the way toward unbiased automation, this study suggests there’s a lot more work to be done.

Here’s the gist: When the LSE team fed identical care records into two popular AI models—Google’s Gemma and Meta’s Llama 3—only the gender was swapped, but the results were far from identical. While Llama 3 showed no significant gender differences, Gemma had some noticeable disparities.

For example, Gemma’s summaries for men tended to focus more on their physical and mental health issues, using terms like “complex medical history” and “poor mobility.” But when the same case notes were inputted for women, the AI described their conditions as “manageable” and, at times, “independent,” even though the medical facts were the same. The result? The model effectively downplayed women’s health needs in comparison to men’s.

The stakes are high when it comes to using AI in healthcare, particularly in long-term care. The care a person receives is often based on perceived need, and when AI tools inadvertently minimize the severity of health conditions—especially for women—it could impact how much care they receive. This isn’t just an academic issue; it’s a potential reality for many people relying on AI tools in their care planning.

Dr. Sam Rickman and the LSE team are pushing for transparency and rigorous testing of AI models in healthcare settings. The recommendation? Mandatory checks for bias in AI systems used in long-term care to ensure fairness across all demographics. This study highlights just how important it is for regulators to step in and ensure that these tools don’t unintentionally perpetuate inequality.

AI can undoubtedly help streamline healthcare administration, but it’s clear from this research that we need to be vigilant about how these tools are being used. If these AI models are going to play a larger role in decision-making, they need to be as unbiased as possible.

For those interested in the full research, you can read the published paper here and access the code used to evaluate gender bias in LLMs here.

This study serves as a reminder that AI is not a magic solution, but rather a tool that needs constant evaluation and refinement. If AI is going to be deployed in life-impacting sectors like healthcare, we have to ensure it’s working for everyone—without reinforcing existing biases.

💼The robot's are now interviewing

It’s official: AI is now conducting job interviews, and guess what? The candidates are not thrilled about it. In fact, many are pushing back hard, refusing to sit through robotic screenings. The latest trend, covered by Fortune, sees job seekers rejecting AI interviews outright, calling the experience dehumanizing and a clear red flag for company culture🚩

Here’s the thing—HR teams love AI. With staffing shortages and a flood of applicants, it’s become the easiest way to sift through thousands of resumes. But for many job hunters, the rise of AI in the hiring process feels like yet another indignity in an already soul-sucking job market💻

So Imagine this (although it's actually reality): You’ve spent hours tailoring your resume, prepped for the big interview, only to log into Zoom and be greeted by a cartoonish avatar with a robotic voice. Yeah, that’s happening😅

Job seekers, like seasoned editor Debra Borchardt, describe the experience as “awful,” with the interview quickly descending into a series of monotonous questions that feel more like a test than a conversation. “I clicked exit after 10 minutes,” she says. “I’m not going to waste my time talking to a machine.”

Other job seekers, like technical writer Allen Rausch, were less than impressed with the lack of human interaction. Despite enduring AI interviews several times, he says they feel like "a cost-saving exercise more than anything" and leave him questioning the company's culture. If HR can't even take the time for a human interaction, what's the company's investment in its people? Fair point.🤔

While candidates are rejecting AI interviews as impersonal, HR teams are doubling down on the tech. With companies flooded with applications, AI interviewers allow HR to quickly filter out candidates who aren't a good fit—leaving the human interviewers with more time to focus on top-tier prospects. Priya Rathod, a workplace trends editor at Indeed, says this is especially common in high-volume hiring sectors like retail and customer service🛍️

But here's the catch: AI can assess objective skills, but when it comes to evaluating cultural add or the nuances of human interaction, it's pretty far behind us humans. As Adam Jackson, CEO of Braintrust, a company distributing AI interviewers, admits, AI "wouldn’t even try" to assess cultural fit. So, are we sure we're not just automating away empathy and personality in the hiring process?🤖

So, if you’re a job seeker, you might just want to double-check if your next interview is with a human before you dive in. And if you're an employer, maybe it's time to ask: does your AI hiring process really reflect the values you want to showcase to future employees? After all, who wants to join a company that treats its candidates like robots. Well, unless they're hiring robots.😑

For a deep dive into the world of AI hiring interviews, check out the full Fortune article here and read more about AI's role in hiring here.

📈GPT-5📈

OpenAI has just launched GPT-5, its most advanced AI model to date, promising smarter, more competent responses and fewer pesky “hallucinations.” In other words, it's like having a PhD-level expert at your fingertips. It’s better at writing, coding, and even translating complex medical jargon into plain English. But while it sounds like a perfect upgrade, the rollout has been anything but smooth.😢

OpenAI made a bold move by removing access to all previous GPT models when GPT-5 launched. This meant that users who had spent months crafting workflows around earlier versions like GPT-4o were suddenly left with a new model they hadn’t prepared for. Cue the outrage: Reddit was flooded with complaints from frustrated users whose workflows broke overnight. Some likened it to losing a trusted collaborator, while others felt that OpenAI had overestimated their loyalty to GPT-5.😯

On paper, GPT-5 sounds fantastic. It can switch between tasks like coding a game or writing a paper with ease, and it’s smarter than its predecessors. But for many users, it’s been a bit too efficient. Descriptions of the model as “abrupt” and “cold” have flooded social media, with some saying it’s like talking to a no-nonsense, overworked secretary instead of the engaging, thoughtful AI they’ve come to love. It’s efficient, sure—but is it too much so?😤

The backlash was loud enough that Sam Altman, CEO of OpenAI, was forced to apologize publicly and announce fixes. GPT-4o will be reintroduced for Plus users, rate limits for GPT-5 will be doubled, and OpenAI will offer better transparency about which model is being used for each query. The good news is, OpenAI is listening and responding quickly to user concerns—proof that the future of AI, while not without bumps, is still on the right path🖥️

📉 So what?

GPT-5 holds a lot of promise, and with the tweaks OpenAI is making, things are likely to settle down soon. However, the rough launch serves as a reminder that even with cutting-edge technology, the user experience should never be overlooked. Maybe next time, OpenAI should ask GPT-5 itself how to launch a product—if anyone’s qualified to do it right, it’s their own AI!

Read more here

🔋Tesla for your energy bills🔋

Tesla—best known for its electric cars and, apparently, its fondness for shaking up industries—has just thrown its hat into the UK energy ring. The company has officially applied for a license to supply electricity to British homes, taking aim at the long-standing dominance of big energy providers. If the UK's energy watchdog, Ofgem, gives the green light, Tesla could be offering power to homes and businesses across England, Scotland, and Wales by next year. Sounds like a big shift, right? But before you start imagining your home powered by Elon Musk’s futuristic vision of solar-powered goodness, let’s break down what’s really going on here.🪫

While most people think of Tesla as the car company that made electric vehicles (EVs) cool, the company has been quietly building an empire in energy. Tesla’s solar energy products and battery storage systems are designed to work in tandem with its EVs, so it’s already got the infrastructure in place to not just sell cars, but to supply and store power. Tesla even runs a power supplier business in Texas, where it helps EV owners charge their cars cheaper and pays them for feeding excess energy back to the grid🚗

So, why the UK? Tesla has sold over a quarter of a million EVs and tens of thousands of home storage batteries in the country. With a sizable customer base already in place, it’s not hard to see why Tesla thinks it’s ready to bring electricity to your home. But then again, Tesla’s EV sales in Europe have dipped recently, with the UK seeing a nearly 60% drop in Tesla registrations at points this year. So, this new venture might just be a way for Musk to recharge his business in the UK📈

🧠So what?

Tesla's entry into the UK energy market could challenge the dominance of traditional energy firms. However, Musk’s political ties and previous controversies could make it a tough sell to some UK consumers. While Tesla promotes an integrated energy model—combining solar, EVs, and batteries—trust could be an issue, especially given the company's past quality control struggles.

💻Intel💻

Intel’s ongoing drama under CEO Lip-Bu Tan isn’t just about chip-making—it’s becoming a prime example of the kind of chaotic, unpredictable political maneuvering that’s become a hallmark of the US tech CEO skillset since Jan 2025.

US president Donald Trump publicly demand for Tan’s resignation and that isn’t a standalone event; it’s more of a rerun of the tactics we've seen before. From trade wars to corporate pressure, this is the kind of political micromanaging we’ve witnessed across the wider Economy: blunt force, questionable diplomatic finesse, and a tendency to shift positions based on whim or political convenience.

Just days after calling for Tan’s resignation due to alleged China ties, Trump meets with the Intel CEO, calling him a success story. Confused much?😯

We're not adept at the 4D chess, but Tan's situation is a microcosm of the corporate-politico whirlwind that companies are often forced to navigate when such unpredictable influence is at play.🤔

It’s unclear whether this meeting will lead to real changes for Intel or just another temporary calm before the next political storm. What’s certain, however, is that Intel, like other tech companies before it, is caught in the middle of a volatile intersection between corporate ambitions and the whims of a president whose leadership style often seems to resemble a poker game—where every move is about keeping everyone guessing🎰

So, while Tan’s future is on shaky ground, the real story here might be about how US businesses have learned to survive—or at least stay afloat—amid the kind of chaos that’s become business as usual. Leets situate this. It is Extremely rare for the President of any country to call for the resignation of a Major CEO on what appears to be a whim👀

🧠So what?

The game is probably about US investment. A successful Intel is good for the USA and, Well Just ask Tim Cook, what one needs to do to get back in the good books. Unfortunately for Intel, Tim committed to spend more than 4x of Intel's entire market cap to do it😅

💻A huge data breach💻

A Salesforce-related data breach has exposed a massive trove of sensitive data, impacting global giants like Google, Louis Vuitton, Adidas, Cartier, and many more💸

While the breach did not involve direct vulnerabilities in Salesforce, it underscores the growing threat of social engineering tactics and third-party app abuse. In particular, Google confirmed that some of its Ads customer data was compromised, including business contact info and related notes, though payment information remained safe🤔

The breach, attributed to the ShinyHunters group, exploited a weakness in Salesforce’s "Connected Apps" feature—an app integration tool that allowed attackers to gain unauthorized access to user data. Through a series of social engineering attacks, the hackers used voice phishing (vishing) to manipulate employees into granting malicious OAuth tokens that enabled them to access vast amounts of personal and business information. The attackers then extracted data through trojanized versions of Salesforce’s legitimate Data Loader tool, bypassing multi-factor authentication and creating a chilling reminder of how much damage human error can cause😳

While the stolen data from Google Ads did not include sensitive financial details, the breach exposed millions of contact records, loyalty program information, and other valuable business data. Other affected companies, such as Chanel, Tiffany & Co., Qantas Airways, and Cisco, reported similar exposures of customer profiles and internal business data. These breaches are a direct result of the increasingly sophisticated methods used by threat actors like ShinyHunters, who have historically targeted high-profile companies for extortion🙆

🧠So what?

What sets this breach apart from others is the use of social engineering tactics to bypass technical security measures. In this case, attackers didn’t exploit Salesforce’s infrastructure; instead, they exploited the humans interacting with it. By tricking employees into approving malicious apps via simple phishing tactics, the attackers gained API-level access to customer data.

This attack shows how even well-protected platforms like Salesforce can become a vector for cybercriminals, as long as human error remains a vulnerability. The growing use of OAuth applications and reliance on cloud-based CRM systems only amplifies the risk, highlighting the need for businesses to adopt stronger user verification protocols.

📈 The tools behind the tech📉

📦Product📦

• GPT pricing

📏Design📏 

• Opticians sans
• Periodic table of System design
• Writing a good design document

👩🏿‍💻Code👩🏿‍💻

• The current state of LLM development

• A Python library for extracting structured information from unstructured text using LLMs
  

🏢The business behind the tech🏢

• An engineers perspective on hiring?

AI for everyone

Boost your career with free AI training – Apply now!
Colorintech is offering free AI courses to Everyone as part of
the Google.org AI Opportunity Fund: Europe. 

This course will equip you with the necessary knowledge and practical experience to succeed in today’s AI driven job market. 

If you’re looking to upskill in AI, boost your employability, and
open doors to exciting career opportunities, this is your chance! 

Applications are open now — visit our website to apply 

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

BTF

We announced DJ AG as one of our headline acts to look out for on this year's lineup

Here is more info: joinbtf.com

Progress in public

Learn about Delivery Manager, User Researcher, and Product careers in the Civil Service

On Aug 20th, we’re teaming up with the awesome team at Government Digital and Data for the second time to talk all about careers and opportunities for:

🌟Delivery Managers

🌟User Researcher

🌟Product Manager

To help with this, we’ll have three inspiring civil servants to share their career journeys and tips for getting a role in the public sector. We’ll be welcoming:

Nicola Hancock - Senior User Researcher, MOJ

Mac-Donnel Graham-Douglas - Product Manager, FCDO

This is a great opportunity for anyone considering a career in the civil service to learn more about the job opportunities available, and how to successfully navigate the hiring process!

Check out the key details below: 

Date: 20th August 2025

Time: 12:00 - 13:30

Where: Zoom

If you’re interested in dialling into this webinar, then be sure to get a space using the link below: 

https://lu.ma/progressinpublictwo

😃What we are consuming😃

💻 GitHub's CEO stands down

😣A taxonomy of LLM ‘hallucinations

🏎️ Why apple may like F1