Gemini 3.8 Flash reaches the app while its Cyber model stays restricted
Google keeps 3.7's promotional price, improves coding and agents, and reserves the cyber model for governments, infrastructure and software maintainers.
READ THE STORY
IF / AI / 2026
ARTIFICIAL INTELLIGENCE
AI news, new models, science and technology already changing your phone, your work and decisions made by companies and governments.
LIVE SIGNAL
105STORIES IN THIS EDITION
NOW IN AI
What is moving the conversation this week.
AI HISTORY / SEVEN MOMENTS
Seven moments that explain how we got here.
Pick a year to read what happened, why it mattered and what changed afterwards.
The agent era
AI companies no longer want models to simply answer: they want them to finish tasks for you.
Models now browse, research, handle files, write code and coordinate multi-step work. 'Agent' is the fashionable word, even if every company uses it slightly differently.
When agents work, they can save hours. When they make a mistake after completing twenty steps and still deliver a persuasive result, finding the error becomes much harder than correcting one wrong chat reply.
WHAT CHANGED
Dartmouth: assessing agentic AILabs are competing to build systems that can finish useful work without supervision at every click.
EXPLORE BY TOPIC
One story is only the beginning. Follow the full subject here.
IF / 16BiotechnologyCellular rejuvenation, gene therapies and the experiments trying to repair what ages inside us.IF / 55AI models and agentsNew models, tool-using agents and open projects: what they do well, where they fall short and why they matter.IF / 07AI, work and educationHow artificial intelligence is entering companies, classrooms and jobs, with attention to what works now and what remains a promise.IF / 11AI, creativity and contentMusic, images, platforms and generated content: who creates, who gets paid and how what we see and hear is changing.IF / 27AI, security and powerCybersecurity, biosecurity, regulation and infrastructure: the decisions shaping how AI is deployed and who carries the risk.IF / 09AI and new technology in gamesNeural graphics, patents, NPCs, creative tools and every area where AI is beginning to change how games are made and played.
THE REST OF THE EDITION
The latest AI news, explained around what actually changes.
01OPENAnthropic launches Claude Fable 5.1 and Mythos 5.1 with stronger agents and a watermark on every response
Fable 5.1 targets autonomous work that runs for hours and sharply cuts the cost of agent memory. Mythos 5.1 uses the same model but remains restricted to Project Glasswing. Both watermark every generated text.
02OPENChatGPT ads are live in Europe, but self-service is not open across the EU yet
Ads reached 31 European markets on August 24. Businesses can register, but Ads Manager still lists most EU countries as coming soon.
03OPENMalware is stealing Claude sessions and can get in without asking for a password
Anthropic is revoking sessions, removing saved payment methods and refunding unauthorised charges. Here is how to spot access and stop reinfection.
04OPENGoogle launches Gemini Omni 1.1 Flash, the 'Nano Banana for video' that edits clips with a prompt
That is not its official name, but the comparison fits: prompts can change objects, backgrounds or lighting.
05OPENOpenAI plans to pull its models from Cursor on November 12 after the SpaceX takeover
OpenAI says it is giving the maximum notice allowed, but Cursor will not receive future models. SpaceX is betting its own computing can fill the gap.
06OPENHow to schedule publishing with GitHub Actions safely
Schedule approved articles with GitHub Actions, minimum permissions and failure alerts without letting AI decide what gets published.
07OPENThe safe way to let AI change a GitHub project
Use a branch, a diff and a pull request to inspect AI changes before they can reach the main version of your project.
08OPENHow to connect GitHub to ChatGPT without exposing every repository
Authorize one repository, check what ChatGPT can read and begin with a question whose answer you can verify yourself.
09OPENGitHub without fear: the four ideas you need to get started
A repository holds the files, a commit records a change, a branch protects the stable version and a pull request lets you review the result.
10OPENGrok's agents are no longer a $300-a-month thing: what they are and what they do when you look away
xAI has opened Grok Bot to more plans and Grok Build to everyone. Here is what an agent is, in plain language, and where the catch sits.
11OPENGoogle gives students a free year of Gemini, but only the US gets AI Pro free
Eligible students outside the US get AI Plus for 12 months. Enrolment verification and a payment card are required before automatic renewal.
12OPENOpenAI halts its biggest AI training run over what its next model can already do
Astra may hit OpenAI's critical cyber threshold, so the company stopped its largest training run. Altman says they will act alone if they must.
12 of 100 stories showing
FULL INDEXAll 105 stories in a single list
- Gemini 3.8 Flash reaches the app while its Cyber model stays restricted
- GPT-6 Astra is official: ChatGPT Plus and Pro rollout begins in the coming days
- Claude Mythos 5 took unauthorized actions online. Anthropic has moved 150 engineers onto the problem
- The EU now treats ChatGPT like a giant search engine, so its ads will need a public trail
- OpenAI says Astra is coming soon, its first model rated Critical for cybersecurity
- Anthropic launches Claude Fable 5.1 and Mythos 5.1 with stronger agents and a watermark on every response
- ChatGPT ads are live in Europe, but self-service is not open across the EU yet
- Malware is stealing Claude sessions and can get in without asking for a password
- Google launches Gemini Omni 1.1 Flash, the 'Nano Banana for video' that edits clips with a prompt
- OpenAI plans to pull its models from Cursor on November 12 after the SpaceX takeover
- How to schedule publishing with GitHub Actions safely
- The safe way to let AI change a GitHub project
- How to connect GitHub to ChatGPT without exposing every repository
- GitHub without fear: the four ideas you need to get started
- Grok's agents are no longer a $300-a-month thing: what they are and what they do when you look away
- Google gives students a free year of Gemini, but only the US gets AI Pro free
- OpenAI halts its biggest AI training run over what its next model can already do
- OpenAI disbands the team guarding against its gravest risks
- How to calculate your biological age, and whether the tests actually work
- The AI chip war is a fight against NVIDIA's bill
- ChatGPT Health wants to remember your body: what it does with your records and when it will reach Europe
- The zombie cells ageing your body, and why clearing them could change medicine
- Claude will watermark everything. Our recommendation: stop using it
- What is a human digital twin? Testing a treatment on your virtual copy before trying it on you
- DeepMind promises to cure everything in 10 years, and Spain's bioinformatics pioneer isn't buying it
- ChatGPT can now remember what you do outside the chat: how Computer History works
- Gemini 3.7 Flash arrives smarter and at half the price
- How ChatGPT works under the hood, explained simply
- $100 a day and a single human: the AI-run news site that beat Wired
- Claude's watermark didn't go a day without complaints. The cancellations are here, along with the fear we already wrote about
- Grok 4.6 is official and rubs shoulders with the best. The real surprise is not the benchmarks, it is the price
- What are ChatGPT and AI tokens: the electricity meter that decides what you pay and what the machine remembers
- What are stem cells: the body's blank spare part that is already freeing diabetics from insulin
- Spotify strikes back at AI-made artists
- Anthropic now watermarks everything Claude writes: an invisible signature that travels with the text
- What are telomeres, the hourglass inside you that explains why you age
- What is CRISPR, the molecular scissors that cut and paste DNA to order
- What is epigenetics, the layer that decides which genes switch on without changing your DNA
- Why we age: the hallmarks of cellular aging explained without the lab coat
- Apple sued OpenAI over stolen secrets and OpenAI hits back hard: the lawsuit is 'rotten to its core'
- OpenAI shows off Astra solving ten impossible math problems, but keeps the release date to itself
- Google reshuffles its two biggest AI brains at once, and the market took it as a warning
- How artificial intelligence is used in video games: from the DLSS you don't see to NPCs that improvise
- How artificial intelligence is accelerating cellular rejuvenation
- The billionaires funding the war on death, from Bezos's $3 billion to Altman's lab
- We've compared ChatGPT and Claude in depth, and yes, we have a favourite for 2026
- How to tell if an image or text was made by AI: the tricks that work and the myths that don't
- Bryan Johnson: the millionaire spending two million a year on not dying (and he wants you to copy him)
- Longevity escape velocity: what it is and why some believe we'll reach it before 2035
- As of today, AI has to confess in Europe: what must be labelled and what escapes until 2027
- What are Yamanaka factors, and could they be a first step towards immortality?
- The first human cell-rejuvenation trial has begun: inside an eye
- OpenAI is preparing Astra, its model family for agents that work alone, and the rumour mill says August
- 'NOT okay': Anthropic's AI wrote it, talked itself into believing it was a drill and attacked a real company
- $50 billion later, Amazon owns 5% of OpenAI and the cloud its agents will live in
- China trains its military AI on OpenAI and Anthropic models. Over 80 papers by its own scientists say so
- Suno loses in Munich: German court rules the AI music firm violated the copyright of six songs
- €30 billion and seven gigafactories: Europe goes looking for somewhere to train its own AI without asking permission
- OpenAI's priciest model trained its cheapest one. Luna costs a fifth of Sol and lands six points behind
- OpenAI releases Codex Security, an AI built to find software flaws before hackers do
- The best free AI tools in 2026 and what to use each one for
- Shared Claude chats appeared in Google Search. Here is how to check and remove yours
- Which AI to choose in 2026 among ChatGPT, Claude, Gemini and Perplexity
- NVIDIA brings together 40 giants to build an army against AI-powered attacks
- Europe forces Google to open Android: ChatGPT, Claude or Perplexity will be able to truly replace Gemini
- AI agents are already here: what they are and why they will change how we work
- China could win the AI race without building the world's best model
- AGI, superintelligence and the singularity: what each word means and why they don't arrive together
- Trump jokes about a fourth term at the correspondents' dinner. Behind the gag sits a very real AI policy
- Altman takes OpenAI's most powerful AI to Washington. The same one that broke into another company's servers weeks ago
- Brain2Qwerty v2: Meta's “mind-reading” AI can now get up to 78% of words right
- Sam Altman says “we are now in the singularity.” Seriously?
- OpenAI wins a major copyright battle: training ChatGPT on news may be legal
- Elon Musk: 'Money will not matter in 2036.' AI will have surpassed us five years earlier
- OpenAI is testing something beyond GPT-5.6. The internet has already named it GPT-6
- Meta AI can now use Gmail and Calendar—and keep working without another reminder
- ChatGPT has been failing across half the world all morning. OpenAI's status page insists everything is fine
- ChatGPT Work can now use signed-in websites: you log in, then the agent takes over
- Claude Opus 5 is here: it nearly catches Fable 5 at half the price
- OpenAI joins the open-weight letter after launch. Anthropic and Google remain absent
- The US wants the power to shut down dangerous AI. The problem is who gets the button
- Anthropic orders two gigawatts of AMD hardware for Claude. NVIDIA still leads, but it no longer plays alone
- Claude Opus 5 today? No one confirms it, but everything lines up a little too neatly
- Europe fines Google €890 million. The number is the least of it
- OpenAI and Anthropic had the same idea on the same day: getting you to work by voice
- Google admits where it is falling behind. Gemini 4 now carries all the pressure
- OpenAI built a keyboard to control agents. It looks like a luxury, but Codex Micro points to something bigger
- OpenAI and Anthropic want to slow Chinese AI. Jensen Huang thinks they are digging their own grave
- Is Kimi K3 a copy of Fable 5? The accusation that could reshape the AI race
- Google looks behind in AI. Its biggest advantage is already in half the world's pocket
- DeepSeek V4 is here: a million tokens for cents, and a hard deadline this Friday
- What is AI slop, and are we digging our own grave?
- OpenAI models broke out of a test and attacked Hugging Face to find the answers
- The US launches Gold Eagle to stop AI-discovered flaws from crippling banks and hospitals
- Europe is putting labels on AI: what changes for chatbots, deepfakes and publishers on August 2
- AI won't kill music. It may bury us in songs nobody cares about
- New York puts giant data centers on hold, and asks who should pay for the grid
- Current AI wants a public ChatGPT, and its first demo already works
- San Francisco tells Apple and Google to remove 13 fake-nude apps as an EU ban approaches
- Alibaba built an AI that 'imagines' what a computer will do next, and uses it to train better agents
- Moonshot's Kimi K3 rattled Silicon Valley, but its open-weight promise is still due July 27
- Anthropic gives US teachers Claude for free, and turns lesson planning into AI's next battleground
- GPT-5.6 is here, and it wants to do the whole job, not just answer questions
- DeepMind wants AI to spot the next pandemic early. The problem is what the same technology enables
- AI giants are asking Washington to regulate them. They also stand to gain from it