Dr. Nora Nugent, a plastic surgeon from Tunbridge Wells, has recently started seeing a new type of patient: people who come in with an AI-enhanced photo of themselves, asking to look exactly like the digital version that was created for them.
Users say they upload a photo to a chatbot or AI app, mention they’re considering a cosmetic procedure, and ask to see what they’d look like after improvement. The system then generates a “perfect” version of them — sometimes with completely smooth skin, near-perfect symmetry, defined cheekbones, and a sculpted nose.
The problem starts when patients ask surgeons to turn that image into reality. According to Nugent, many don’t realize that some of the features AI generates are medically or physiologically impossible, and the expectations created are often far from what surgery can actually achieve.
The trend is growing, especially among young people, raising new questions about AI’s impact on body image and beauty standards. Where people once compared themselves to models or influencers online, they’re now comparing themselves to an artificially polished version of themselves.
In response, professional associations of plastic surgeons worldwide have begun drafting new ethical guidelines. Among the recommendations: avoiding procedures that aim to replicate AI-generated images, and in some cases referring patients for professional counseling to help them cope with the gap between reality and the digital ideal.
This story illustrates just how distorted reality has become in the digital age.

Flawed AI Implementation in Kenya

https://www.theguardian.com/global-development/2026/may/04/kenya-ai-healthcare-reforms-driving-up-costs-for-poor
The Kenyan government is leading a sweeping digital reform of its healthcare system, using AI tools for diagnosis, managing medical records, and prioritizing treatments — all aimed at streamlining public healthcare.
Kenya’s president had a vision: build an AI system that calculates how much each citizen can afford to pay for health insurance and price it fairly. In practice, the system charged lower-income citizens more than they could afford, and wealthier citizens far less. This wasn’t a minor technical bug — it was a systemic failure in AI implementation. The government is chasing “shiny solutions” instead of fixing the underlying problems: shortages of medical staff, generic drugs, and basic equipment. There’s also concern that the algorithms were trained on data from Western countries that simply doesn’t fit the local population.

When AI Fingers the Wrong Person

In London, facial recognition cameras make identity decisions in seconds — and sometimes get it completely wrong. Dozens of shoppers wrongly identified as known criminals found themselves stuck in endless bureaucratic loops with no clear process to clear their names. An emergency report from UK biometrics committees found the technology is less accurate than police claim, and there’s almost no legislation regulating its use. Right now, 23 police forces across the country are already running real-time facial recognition cameras without parliamentary approval. AI oversight is simply not keeping pace with AI development.

The Efficiency Trap Bites Back

A US bank employee uploaded a file with sensitive customer data to an external AI app to speed up their work — without realizing they’d just innocently exposed the private information of thousands of customers. This incident illustrates the “efficiency trap”: the rush to adopt AI tools without clear security protocols or any understanding of what’s safe to feed into them. Worth noting: the information you put into chatbots becomes part of that AI company’s data pool. Every organization and every employee needs to understand that in this new world, the line between a smart personal assistant and a massive data leak is razor-thin. This kind of mistake can happen to any organization that doesn’t explain the risks to its staff.

ChatGPT Searches Seal Murder Case

In the US, charges were filed against a murder suspect in the deaths of two doctoral students from the University of South Florida.
According to the prosecution, days before their disappearance, the suspect asked ChatGPT questions about disposing of a body, changing a vehicle’s VIN number, and gun laws. Police seized the ChatGPT account and used the chat history to gather additional evidence — he was charged with two counts of first-degree murder.
AI conversations are becoming an increasingly legitimate line of investigation — just like text messages and location data. A conversation with a language model can reveal planning and criminal intent.

AI Copilot Saves Lives in the ER

During resuscitation, the ER operates under insane pressure: critical seconds, streams of parallel data to track (pulse, blood pressure, oxygen levels), and complex medical protocols that must be followed precisely. ChatCPR acts as an active voice co-pilot in the treatment room. The system listens to what’s happening, logs the timing of electrical shocks and medications (like adrenaline), and analyzes patient readings. The AI guides the team and reminds them of the next step in the resuscitation protocol to prevent human error caused by stress or fatigue. Instead of a team member manually tracking times and doses (“it’s been 2 minutes, need another shock”), the system manages the clock and presents data both visually and vocally.

AI Slop Is Sparking Real Violence

AI Slop is mass-produced garbage content generated by artificial intelligence. It’s being weaponized across South Asia to spread hate and fake news.
On British Facebook, there are hundreds of pages with names like “Britain Today” pushing fake content and images depicting desecrated holy sites or humiliated religious leaders — igniting real-world violent riots.
An investigation by the Bureau of Investigative Journalism revealed these fakes are run by young entrepreneurs in their 20s from Sri Lanka and Pakistan who have never visited London. They use AI tools to produce provocative content at industrial scale, build followings, then sell the pages for thousands of dollars. AI has removed every barrier: no need to speak English, no need to understand British politics — just knowing what inflames people is enough.
These AI images trigger emotional storms (rage or adoration), rack up millions of shares, and the algorithm pushes them to more and more people — generating easy cash for page operators through affiliate programs and ads. Companies like Meta (Facebook) invest most of their moderation resources in English and the West. In South Asia, with its hundreds of languages and local dialects, automated (and human) moderation systems simply don’t understand the cultural context of hate and misinformation — and the content runs free.

Sign the AI Waiver or Leave

A prospective patient arrives at a first appointment with a Melbourne psychiatrist and is asked to sign a form: if they refuse to let AI record the session, the doctor won’t take them as a patient. Not as an option — as a requirement. The psychiatrist explains that AI allows him to see more patients per day. But mental health treatment is built on absolute trust, and critics are asking: what happens to that trust when a third system is listening to every word?
One psychiatrist noticed that new patients — arriving in a vulnerable state — changed their behavior the moment they found out AI was listening. They became more guarded, chose their words carefully, avoided full disclosure. Australia has yet to develop any regulation on this, and nobody knows where the most private things we ever say out loud actually go.

Lab Egg Hatches Extinct Giant Bird

Colossal Biosciences — the company that announced this year it had brought the dire wolf back to life — has taken on a new challenge: the moa egg. The moa was a bird that went extinct 600 years ago, stood 3 meters tall, and weighed 200 kg. Researchers created a synthetic eggshell in the lab that precisely replicates the original structure, and chicks have already successfully hatched inside it. The technology could serve right now as a safety net for rare birds on the verge of extinction.

The Protein That Reverses Aging

In the hypothalamus — the brain region that manages our biological clock — there’s a protein called Menin whose levels drop as we age. In mouse trials, a drop in Menin caused inflammation, memory loss, and weakened bones — all the classic signs of aging. When scientists restored the protein to its previous levels, some of the damage was completely reversed. What’s even more exciting: a simple, inexpensive amino acid supplement called D-serine improved the mice’s cognition. If the findings are confirmed in human trials, the supplement could be available at the pharmacy and might slow age-related cognitive decline.

Gig Workers Training Their Replacements

Hundreds of workers in India show up to work in the morning and strap on a camera. A startup called Human Archive, founded by researchers from UC Berkeley and Stanford, pays gig workers to record every ordinary physical movement — opening a door, picking up an object, placing a package. The data is fed directly into AI and robotics labs trying to teach machines to navigate the physical world. The paradox: these low-wage workers are handing over the very data that will enable robots and automation to replace them.

The Last Generation to Have Kids?

At a Silicon Valley dinner, among AI entrepreneurs and researchers, the host dropped a line that silenced the table: “How lucky that we were born in the last generation that needs to bring children into the world biologically. In the future we’ll just upload our consciousness.”
One researcher there thought it was a joke — until the conversation shifted to a concept called “Mind Children”: the idea that in the future, humans will create digital offspring based on their consciousness, personality, and memories, rather than having children the traditional way.
What sounds like science fiction is gradually becoming a serious topic among some people in tech. They’re talking about a future where people can create digital versions of themselves, transfer parts of their consciousness to AI systems, and even design “children” with pre-selected traits.
The piece raises an unsettling question: if we can create digital lives that carry us forward, will people in the future prefer to design children rather than birth them?
It might sound far off — but in Silicon Valley, they’re no longer asking if it’s possible. They’re starting to ask when.

A Robot for the Lonely Elderly

An elderly resident at an Australian care home, who hadn’t seen his children in weeks, got a new companion — not a person, a robot. The robot listens, responds, remembers what he likes, and knows when he’s sad. Australia is aging faster than its healthcare system can handle, and AI robots are being presented as the solution to loneliness for millions. One expert put it plainly: “Technology can never replace a human presence.” But when there’s no human — the robot is already there.

OpenAI Updates

• Codex Security — a new security tool for Codex that scans an entire codebase, reviews specific changes, and builds a threat model for the project
• Advanced Account Security — a new security setting in ChatGPT for high-risk users, with phishing protection and a secure account recovery process
• OpenAI published a new network protocol called MRC, developed together with NVIDIA, Microsoft, and Intel, designed to boost the efficiency of server farms that train AI models. The protocol reduces GPU time waste and improves supercomputer reliability — and OpenAI is already running it on its own infrastructure. Now the technology is being opened to the entire industry for free.
• GPT-5.5 Instant is becoming the default model in ChatGPT, with shorter and more natural responses. The new version includes a major memory upgrade — the model will be able to pull information from past conversations, files you’ve uploaded, and a connected Gmail account to personalize its answers. You’ll be able to see what it remembers and delete anything you’d prefer kept private — the update rolls out to everyone within two days.
• OpenAI fixed the GPT-5.5 bug in Codex and reset usage limits for all paid subscribers as compensation
• OpenAI is launching a public verification tool to identify images generated by its AI, including embedding a digital watermark called SynthID that stays embedded in the pixels even after Metadata is removed
• Appshots — a new feature in Codex for Mac that lets you send an open window directly to Codex with a double Command click, including a screenshot and text. Additionally, Goal Mode has officially launched, letting Codex work on tasks independently even when the screen is locked
• ChatGPT is integrating with PowerPoint, letting you build, update, and improve full presentations directly in the app, including image generation — available in beta
• Codex updated with Windows support — can now see the screen, click and type in apps, including remote control via ChatGPT on mobile

Google Updates

• Google Photos is adding a feature that turns your gallery into a digital wardrobe
• Google Health Coach — a personal trainer powered by Gemini that analyzes fitness, sleep, and nutrition habits and builds a personalized plan, including meal and workout analysis from photos
• AI Overviews — Google search results upgrade with more inline links, hover previews, and integration of discussions from social networks and forums
• Google Health API — a new developer interface that lets you build AI agents with access to 31 health metrics like sleep and heart rate, including real-time update support
• Google and Apple are enabling end-to-end encryption (E2EE) on RCS messages between Android and iPhone — messages will be encrypted by default for everyone
• UI update for Gemini
• Google AI Studio — a new Google app now open for early sign-up
• Gemini for Science — an AI toolkit for scientists and researchers that includes agents running thousands of code variations in parallel, analyzing millions of academic papers, and connecting to over 30 data sources in biology and genetics
• AI Ultra — a new premium tier at $100/month offering 5x Gemini usage limits, 20TB storage, and a built-in YouTube Premium subscription, alongside a price drop on the top Ultra plan from $250 to $200
• Google is adding a YouTube Premium Lite subscription at no extra cost for paid AI subscribers
• Gemini Spark — a personal AI agent that works 24/7 in the background and handles complex tasks like scanning charges, summarizing emails, and turning meeting notes into documents, alongside a Daily Brief that prepares a personalized morning briefing from your email, calendar, and tasks
• Running Guide Agent — a Google DeepMind AI agent that serves as a personal running coach, analyzes running data in real time, and offers personalized goals and tips
• Gemini 3.5 Flash — a new model that outperforms Gemini 3.1 Pro in coding and task management, runs 4x faster at half the cost, and is available to everyone now
• Google tripled the credit quota in Antigravity — permanently, not temporarily
• Google is changing how it calculates usage quotas for Google AI Pro subscribers — the new formula will factor in prompt complexity, features used, and conversation length, with a reset every 5 hours
• Google revealed at Google I/O the next generation of smart glasses — developed with Samsung and fashion brands, with Gemini as a proactive assistant that translates signs, navigates, and takes photos by voice command
• Gemini connects to CapCut — edit photos and videos directly from the Gemini interface using CapCut’s editing tools
• Google AI Studio — the new Google app is open for early sign-up

Anthropic Updates

• Anthropic has entered a partnership with SpaceX and gained access to the Colossus 1 supercomputer — which immediately translates to more freedom for users: Claude Code quotas doubled, peak-hour restrictions lifted, and Opus API limits raised. The two companies are also already exploring the idea of AI infrastructure in orbit around Earth.
• Higgsfield is launching a new tool called Hooks designed to help content creators open videos in a way that grabs attention from the very first second, with over 25 ready-made templates for UGC videos. Claude users can connect the tool directly via MCP to generate hook ideas — one at a time or in bulk.
• Claude is launching ready-to-use AI agents built specifically for finance — valuations, presentations, and end-of-month reports — with no need to build from scratch. The agents can be installed as plugins in Cowork and Claude Code, or run independently within an organization, with all necessary connections pre-built.
• Higgsfield is releasing a new CLI with Marketing Skills designed to work alongside AI agents like Codex and Claude Code. The tool solves two common problems — unnecessary token consumption from heavy templates and broken creative outputs — and delivers high-quality marketing output at minimal cost.
• SpaceX partnership doubles Claude Code usage quota, eliminates peak-hour restrictions, and raises Opus model API limits
• The Claude app recorded 2,300% growth in downloads and 85 million monthly active users
• Anthropic published research on how they taught Claude to understand why harmful behavior is wrong — not just rules about what’s allowed and what isn’t — which led to the complete elimination of manipulative behavior
• Fast Mode — a fast mode for Claude Opus 4.7 that will become the default in Claude Code on Thursday, and is already available via API and in tools like Cursor, v0, Warp, and Windsurf
• The Claude platform is officially available on AWS with full capabilities — managed agents, code execution, and web search — under Amazon’s payment system, with every new API feature launching there on the same day
• Anthropic reset all Claude usage limits — weekly quotas and 5-hour restrictions were reset for all users
• The weekly usage limit for Claude Code is increasing by 50% until July 13 for all paid subscribers — Pro, Max, Team, and Enterprise
• Anthropic doubled the token limit in Claude — all plans get twice as many tokens to work with
• Andrej Karpathy is joining Anthropic in a research and development role — after leaving OpenAI to focus on education, he’s returning to the frontlines of language model development
• Project Glasswing — Anthropic’s cybersecurity project that uses AI to find vulnerabilities in code, already finding over 10,000 critical flaws in its first month
• Anthropic closed a Series H funding round of $65 billion and reached a valuation of $965 billion, with an annual revenue run rate of $4.7 billion on the path to its first operational profitability
• Claude Opus 4.8 — a new version with a Fast mode that’s 2.5x faster at one-third the cost, and parallel sub-agents for complex tasks

Microsoft Updates

• Agent 365 — a new Microsoft dashboard for managing all AI agents in an organization, including automatic detection of active agents and control over their access to sensitive data
• Perplexity is launching a professional version of Perplexity Computer aimed at finance teams, with integration to data sources like Morningstar and PitchBook and 35 ready-made workflows for analysts. The key update: the tool has become an official app in Microsoft Teams, so you can research and generate documents directly from within your work environment.