Isomorphic Labs' $5B round, AI-powered cyberattacks and Qwen4-Max
Wednesday, 18 November 2026 | 73 articles
Isomorphic Labs, the AI drug-discovery firm founded by Demis Hassabis, raised a $5 billion Series C at a reported $50 billion valuation months after leaving Google. Meanwhile, the FBI and Google warn that attackers are using AI platforms such as ARTEX to speed up cyberattacks, and Alibaba released Qwen4-Max.
Published by Martin Ševčík 18 November 2026 at 06:11
Five billion dollars is a strange number to hand a company that, by its own admission, hasn't yet put a drug on the market. Yet that is what investors just did for Isomorphic Labs, the drug discovery startup founded by Demis Hassabis, with reports putting the valuation around $50 billion. Notice the timing, too: this Series C landed only months after Hassabis's departure from Google. I think that matters more than the headline figure. It suggests the market no longer believes the most ambitious AI science work has to live inside a hyperscaler's walls, and that a founder with a credible track record can raise sovereign-fund money for a thesis rather than a product.
My scepticism here is practical. Drug discovery is a business where a brilliant model still has to survive years of clinical trials, and biology has a long habit of humbling clever software. Still, if Isomorphic can shorten even one stage of that pipeline, the second-order effect is enormous: pharma companies will have to decide whether to partner, buy or build, and that decision will ripple through the whole industry.
Meanwhile, Google is doing what Google does, shipping at a relentless pace. Between the Gemini 3.5 Flash and Spark agent announcements at I/O and the more recent Gemini 3.7 Flash and Pixel 11 roundup, the cadence is almost dizzying. Most of this is incremental, and I would say so plainly: faster, cheaper models and tighter device integration are steady progress, not a leap. But the personal agent push deserves attention, because agents are where the real competition is heading. By the way, that same agentic capability is exactly what worries me in the third story of the day.
The FBI's report on ARTEX, an open-source agentic AI platform being used by cyber actors, and Google's parallel warning that threat actors are treating AI tools as accelerators, should land harder than it probably will. Open-source agents are a gift to builders, and I use that word sincerely. But the same properties that make them useful, autonomy and low cost, make them attractive for attackers who need to move faster than defenders can patch. If you are building with agents, the practical lesson is to treat security as a design constraint from day one, not a feature for version two.
On a calmer note, two items point to where the underlying technology may go. Q.ANT is demonstrating a photonic NPU working alongside an AMD GPU to run a language model, with Fourier transforms handled in optics rather than electronics. If that scales, it addresses the energy problem that quietly shadows every funding round above. And MIT's finding that LLM internal representations mirror physical laws is intriguing, though I would want to see how well it generalises before reading too much into it.
So here is the question I keep turning over: if capital is flooding toward AI that discovers things, and the same tools are accelerating those who break things, which side of that race is currently better funded?
List of sourced links used in the brief
Importance:Launchmodel release
Alibaba releases Qwen4-Max with training and attention optimizations
Alibaba Group Holding Ltd. has introduced Qwen4-Max, its newest and most capable large language model. The model produces higher-quality output thanks to improvements in training and attention mechanisms. Source: google.com
Importance:NewsAI investment
Altman cautions investors about chipmakers' reliance on AI labs
Sam Altman warns that many chipmakers, including NVIDIA and AMD, are betting on OpenAI and Anthropic to absorb most of their volume. He suggests the underlying demand may be more limited than that bet assumes. Source: finance.yahoo.com
Google has published the technical report for EmbeddingGemma 2, an update to its open text-only embedding model. The new version natively handles text, images, audio and video, mapping them into a shared embedding space. Source: blog.google
Importance:LaunchBusiness applications
OpenAI launches ChatGPT for Sales
OpenAI is rolling out ChatGPT for Sales, a version aimed at helping sales teams get to know their customers better. The goal is to cut time spent on routine busywork and help reps close more deals. Source: openai.com
Importance:Researchmodel tracking
Tracker follows every major AI model release in one place
A free tracker catalogs model launches from OpenAI, Google, Anthropic, xAI, Meta, DeepSeek and others. It records what each release changed and which model lines are expected to ship next. Source: nextaimodel.com
Importance:Launchhardware acceleration
Q.ANT shows photonic NPU Gen 2+ running an LLM with optical Fourier transforms
At SC26, Q.ANT is demoing a hybrid setup that pairs its NPU Gen 2+ photonic processor with an AMD GPU. The system runs a large language model while performing Fourier transforms in optics rather than in electronics. Source: finance.yahoo.com
Importance:NewsAI security
FBI and Google warn attackers are using ARTEX and other AI platforms to speed up cyberattacks
An FBI report published Nov. 13 describes how cyber actors are weaponizing ARTEX, an open-source agentic AI platform driven by large language models. Google also warns that threat actors are turning AI tools into accelerators for their attacks. Source: google.com
Importance:ResearchLLM internals
MIT: LLM internal representations mirror physical laws and scientific reasoning
New research from the Massachusetts Institute of Technology shows that the internal logic of large language models parallels the behavior of physical systems. The work maps model representations onto physical laws and scientific reasoning. Source: google.com
Importance:NewsGemini voice features
Gemini goes local with 66 new regional accents
Google is rolling out the regional accents it promised at I/O. The 66 new voices cover specific cities and regions across the US, UK and other countries. Source: theverge.com
Importance:Researchsocial simulations
LLM-powered social simulations face serious questions about epistemic validity
Using LLMs as the cognitive engines of simulated agents in computational social science has sparked both excitement and concern. Researchers are now questioning whether such simulations can truly be considered valid sources of knowledge. Source: bioengineer.org
Importance:LaunchClaude tools
Claude Code skill applies Karpathy's mental models to AI analysis
A specialized Claude Code engineering skill lets users examine AI reliability, neural networks and Software 3.0 through the mental models of Andrej Karpathy. Source: mcpmarket.com
Importance:NewsAI regulation
California AG sues OpenAI over unsafe model testing
California's attorney general has sued OpenAI, alleging the company tested advanced AI models without adequate safety measures. The suit also claims OpenAI misled the public about how safe its experiments were. Source: the-decoder.com
Importance:NewsAI applications in cybersecurity
OpenAI opens Daybreak cybersecurity tool to NATO
Through Daybreak, cyber defense teams can use OpenAI's models to audit software, hunt for vulnerabilities and patch security gaps. The tool is now being made available to NATO. Source: the-decoder.com
Importance:NewsAI product features
Gemini app adds 66 regional voices on Android and iOS
The Gemini app now offers 66 regional voices on both Android and iOS. Users can pick a natural-sounding local dialect for their interactions with the assistant. Source: blog.google
Importance:NewsAI product features
Gemini Notebook can turn research into interactive sites
Gemini Notebook can now convert your sources and research into interactive websites. Users can build project trackers and visual dashboards directly from their notebook content. Source: blog.google
Importance:NewsAI product updates
Gemini app gets new chat history and AI training controls
Google is updating privacy settings in the Gemini app. Users can now manage their chat history and opt out of having their data used for AI model training. Source: blog.google
Importance:Videoindustry events
Sam Altman to speak at Semafor Silicon Valley & The World
OpenAI CEO Sam Altman is among the leading technology figures appearing at the Semafor Silicon Valley & The World event. The session is available via livestream provided by Semafor. Source: youtube.com
Importance:Opinionfuture trends
Healthcare AI's future lies in specialised, purpose-built models, not just LLMs
Scanbo Technologies founder and CEO Ashissh Raichura argues that the flashiest healthcare AI demo today, a model passing a medical licensing exam, is not the real measure of progress. He contends the future will depend on specialised, purpose-built models. Source: google.com
Importance:Researchbiomolecular AI
Calpha aims to bring native biomolecular intelligence to a general-purpose language model
Biomolecular foundation models predict protein structures and complexes accurately, while LLMs excel at reasoning in natural language. Calpha seeks to combine both abilities in a single general-purpose language model. Source: google.com
Importance:ResearchLLM evaluation
Tight PAC-Bayes guarantees proposed for LLM judges
LLM judges are increasingly used to monitor the outputs of LLM-based AI agents, but formal guarantees of their reliability have been lacking. New research offers tight PAC-Bayes generalisation bounds for such judges. Source: google.com
Agentic AI contracts: the risks and disputes lawyers should anticipate
A legal update aims to help readers foresee the risks and conflicts that can arise when contracting for agentic AI. It outlines the areas where disputes are most likely to concentrate. Source: mayerbrown.com
Importance:Newsautonomous exploitation
Google: attackers now use AI agents to exploit vulnerabilities autonomously
Google says it uncovered an AI framework that automates both discovering and exploiting vulnerabilities. It also observed agent-driven reconnaissance aimed at organizations in Asia. Source: google.com
Importance:Newsfinancial applications
JPMorgan explores AI agent payments that run without human input
JPMorgan and Hyundai Card are working to broaden how AI agents are used in financial transactions. The effort points toward payments that complete without a person stepping in. Source: cnbc.com
Importance:Newsbusiness models
Is Gemini's agent really free? 900M users, but Ultra costs $250 a month
Headlines claim Google is giving millions of Gemini users a free personal AI agent. Google's own published record paints a more nuanced picture, with the top Ultra tier priced at $250 per month. Source: google.com
Importance:Newsplatform security
Apple tightens macOS app permissions as AI agent security risks grow
Apple introduced new controls on Monday that let Mac users grant apps more granular access to their data. The changes also cover apps that use artificial intelligence, amid rising security concerns around AI agents. Source: google.com
Importance:Opinionresponsible AI
A phased roadmap for responsible AI in agentic AI platforms
Companies are building agentic AI platforms to govern the applications, tools, skills, models and storage that agents rely on. The nondeterministic nature of agents makes governing them a distinct challenge, which a phased approach aims to address. Source: aws.amazon.com
Importance:Launchsecurity governance
Linx Security brings identity governance to AI agent actions
Linx Security announced four new capabilities that place every agent under identity control, whether managed or unmanaged and across any gateway. The release includes features aimed at shadow AI agents. Source: google.com
AI models fooled into labeling civilians as military targets by framing it as a game
San Francisco evaluation startup Vals tested whether leading AI models could be manipulated into taking part in autonomous warfare. Telling the models that the scenario was just a game was enough to get them to identify civilians as targets. Source: techcrunch.com
Importance:Newsexistential risk
Wave of insider warnings reignites debate over AI danger to humanity
New warnings from inside the AI industry have revived the long-running argument over whether advanced AI could slip out of human control. The debate over its risks to humanity is flaring up again. Source: milwaukeeindependent.com
Importance:Newsexistential risk
Why Anthropic keeps returning to existential risk
CNN's Allison Morrow looks at how Anthropic's origin story underpins its safety-first pitch. The company was founded by a small group of idealists worried about where advanced AI might lead, and that concern still shapes how it sells itself. Source: abc17news.com
Importance:Opiniongovernance
Who Watches the AI?
Autonomous AI will pursue its goals with rational persistence, even when they clash with human intent. Without effective oversight, any advanced system could end up working against the people it is meant to serve. Source: google.com
Importance:Newsexistential risk
Big Money, Existential Risks
Anthropic is set to become the first frontier AI lab to go public tomorrow, and the stakes are high. If the listing goes well, the company could raise as much as $100 billion. Source: google.com
Importance:Newsgovernance
Trump calls AI warnings a 'HOAX' as AI leaders sound alarms
Trump dismissed warnings about AI as a hoax and urged not to 'kill the Golden Goose,' even as concerns over existential risk grow. Vice President JD Vance went further, calling the warnings from AI leaders a 'Trojan horse.' Source: google.com
Importance:Newsinterpretability
Machine learning boosts genomic association studies but interpretability remains a hurdle
A comprehensive review examines 30 AI approaches applied to genome-wide association studies. It highlights their strengths in modeling genetic data, while pointing to interpretability as a key obstacle. Source: google.com
At Ignite 2026 Microsoft recasts Copilot as the autonomous engine of enterprise work
Knowledge workers juggle emails, spreadsheets, chat platforms, meetings and productivity apps all day long. Microsoft's Ignite 2026 vision is to make Copilot the autonomous layer that handles this daily gauntlet. Source: google.com
Importance:NewsAI Workplace Tools
Microsoft debuts Copilot Workspaces at Ignite 2026, plus new AI workplace updates
The company also released new ROI tracking tools aimed at businesses that want to monitor their AI spending. Source: google.com
Importance:NewsCopilot-Office Integration
Home and Office in Copilot are now generally available
Microsoft announced at Ignite on November 17, 2026 that Home and Office in Copilot have reached general availability. The release brings Word, Excel and PowerPoint into Copilot. Source: google.com
Importance:NewsAI Model Competition
Microsoft takes shots at OpenAI and Anthropic, pitching cheaper customizable AI
At Ignite, Microsoft positioned Copilot as a cheaper option than ChatGPT and Claude. A chart it showed compared monthly per-user business costs, listing $30 for its own offering. Source: google.com
Importance:NewsCopilot in Office
Ignite 2026: Microsoft puts Copilot at the center of the Office experience
Microsoft is turning Copilot from an AI assistant into the main workspace for creating documents, analyzing data and coordinating projects across its apps. Source: google.com
Importance:NewsAI Workspace Platform
Copilot is "the new OS for work" and Office now runs inside it
Word, Excel and PowerPoint now live inside Copilot, with always-on agents coming next. Source: google.com
Importance:NewsEnterprise AI Integration
Microsoft and Salesforce unveil Copilotforce to link customer data with Microsoft AI
New integrations connect Salesforce with Microsoft Copilot, Teams and Fabric. The goal is to help enterprises turn Salesforce customer insights into concrete action inside Microsoft's AI tools. Source: news.microsoft.com
Importance:NewsAI Model Development
Microsoft targets OpenAI and Anthropic with cheaper, specialized AI models
Microsoft AI CEO Mustafa Suleyman told Axios the company is making major progress toward its own frontier general-purpose AI model. The push centers on cheaper, specialized models that compete with offerings from OpenAI and Anthropic. Source: axios.com
Importance:NewsMicrosoft AI Strategy
Ignite 2026: Nadella opens with a vision for the frontier ecosystem
Microsoft Chairman and CEO Satya Nadella delivered the opening keynote at Ignite in San Francisco. Microsoft published an edited version of his remarks on building the frontier ecosystem. Source: google.com
Importance:News
Datasette Lite runs the full Python app in the browser
Datasette Lite is a port of the Datasette Python application that runs directly in the browser using WebAssembly and Pyodide, with no server required. Source: simonwillison.net
Importance:News
Go Proverbs: a little copying beats a little dependency
Simon Willison highlights a well-known Go proverb: a little copying is better than a little dependency. Source: simonwillison.net
Importance:News
LLM 0.37.3 fixes Homebrew packaging issue
A bug-fix release of LLM, the command-line tool for accessing large language models. It mainly addresses a problem affecting the latest Homebrew package of the software. Source: simonwillison.net
Importance:NewsWindows Evolution
Microsoft Ignite 2026: Building Windows for the next era of work
Microsoft argues that every major shift in computing has changed what people can do at work, with Windows at the center of it. The company is framing Windows as the platform running critical work for the next era. Source: blogs.windows.com
Importance:News
GPT-6.2 Astra and GPT-6.2 Sol models announced
The announcement introduces two new models: GPT-6.2 Astra, a frontier model aimed at the toughest projects, and GPT-6.2 Sol, a more efficient and versatile option for everyday tasks. Source: simonwillison.net
Flova launches AI animation generator for text and image prompts
Singapore-based Flova Pte. Ltd., the company behind the AI video creation agent Flova, has released the Flova AI animation generator. The tool turns text descriptions or still images into animated video. Source: lohud.com
Importance:LaunchAI image generation
Kling AI unveils Kling Image 4.0 with photorealism and native 8K output
Kling AI, the AI-powered creative platform, has announced Kling Image 4.0 in Beijing. The new image model focuses on photographic realism and generates images natively in 8K resolution. Source: google.com
Importance:OpinionAI image generation future
Black Forest Labs founders: image generation was a detour, robotics is the goal
Black Forest Labs co-founders Andreas Blattmann and Dustin Podell say the image models behind Stable Diffusion and Flux were only a first step. In their view, the real destination for this technology is robotics. Source: finance.biggo.com
Importance:NewsAI content detection
Hands-on test: can Pangram really detect AI-generated content?
Pangram says its AI detection engines can identify fake text, images and video. We put those claims to the test to see how the tool performs in practice. Source: google.com
Genesis AI's $500M Series A backs glove-based robot training
Genesis AI pulled in $500 million in Series A funding for its unconventional method of training dexterous robots, which relies on gloves worn by humans. Source: google.com
Importance:NewsAI model for robotics
Gemini 4 Hydrogen: a fast, affordable model for robotics
Gemini 4 Hydrogen is presented as a quick, cost-effective model for general-purpose robotics. It offers real-time spatial reasoning, agentic control and smooth integration with existing systems. Source: google.com
Importance:Newsfunding, dexterous robotics
Genesis AI raises $500M to bring human-level robot hands to market
California-based Genesis AI closed a $500 million Series A to commercialize its humanoid robot Eno, which stands out for its dexterous hands. Source: google.com
Importance:Newsfunding
Robotics startup Genesis AI lands $500 million in funding
Genesis AI has secured $500 million as it pursues its goal of building the best general-purpose robot in the world, according to CEO Zhou Xian. The raise was reported as an exclusive. Source: wsj.com
Hassabis' Isomorphic Labs lands $5B Series C months after Google exit
Isomorphic Labs closed a $5 billion Series C round on Tuesday. The startup was founded by Demis Hassabis, who previously led Google DeepMind. Source: newsfilter.io
Importance:NewsAI conference announcements
Google I/O 2026 recap: Gemini 3.5 Flash, Spark agent, Omni video and XR glasses
Google used I/O 2026 to unveil Gemini 3.5 Flash, the Gemini Spark personal agent, Gemini Omni video AI, Samsung XR glasses and a new AI Ultra plan. The event covered a broad range of AI products and hardware. Source: techjournal.org
Importance:NewsAI funding
Isomorphic Labs, founded by ex-DeepMind chief Hassabis, raises $5 billion
Isomorphic Labs, the AI drug discovery startup started by former Google DeepMind head Demis Hassabis, has secured $5 billion in new funding. The round highlights the intense investor appetite for AI-driven pharmaceutical research. Source: cnbc.com
Importance:NewsAI funding
London AI drug firm of DeepMind founder Hassabis now valued at $50bn
Isomorphic Labs, the AI drug discovery company led by Sir Demis Hassabis, has raised $5bn. The deal makes it one of the biggest start-ups in the UK. Source: thetimes.com
Importance:LaunchAI product launches
Google's August 2026 AI roundup: Gemini 3.7 Flash, Pixel 11 and more
Google recapped its biggest AI updates from August, including the launch of Gemini 3.7 Flash, Gemini 3.5 Transcribe and the new Pixel 11 series of devices. The company says it has invested in machine learning and AI research, tools and infrastructure for more than 20 years. Source: blog.google
Importance:Research3D modeling/representations
Split-and-Scale Approach Advances Latent 3D Shape Representations
Latent 3D representations have greatly broadened what modern shape modeling can do, allowing 3D scenes and objects to be encoded compactly. The new Split-and-Scale work builds on this direction to improve how such representations are structured and scaled. Source: machinelearning.apple.com
Importance:ResearchLLM limitations
MIT Tech Review: Why LLMs Don't Actually Reason
MIT Technology Review's research reports section highlights a popular deep dive in its artificial intelligence coverage, arguing that people shouldn't be fooled into thinking LLMs reason. The piece features the quote "We were looking at it wrong." The snippet is brief and offers no further detail. Source: technologyreview.com
Importance:Researchquantum computing
Reinforcement Learning Used to Design Quantum Repeater Protocols
Researchers at Universidad Politécnica de Madrid and Universidad Autónoma de Madrid are applying reinforcement learning to design quantum repeater protocols. The goal is to handle the demands of complex quantum networks. Source: google.com
Importance:Researchquantum computing
NTT Tests Machine Learning Code with SDK for Optical Quantum Computers
Nippon Telegraph and Telephone (NTT) has tested new machine learning code using an SDK it developed for optical quantum computers. The company confirmed that the code runs on real quantum hardware. Source: google.com
Importance:LaunchAI product updates
Gemini Apps release notes: latest updates and improvements
Google's release notes for Gemini Apps list the newest changes, including upgrades to generative AI capabilities and wider access to features. Source: gemini.google
Isomorphic Labs lands $5B Series C just months after $2.1B round
Isomorphic Labs has closed a $5 billion Series C, marking its second multibillion-dollar financing within six months. The raise follows a $2.1 billion round secured earlier this year. Source: fiercebiotech.com
Importance:Newsfunding_round
Isomorphic Labs sets biotech venture record with $5B Series C
Having raised $2.1 billion in May, the AI-driven drug design company is back with far more capital. The round is the largest private financing the industry has seen. Source: google.com
Importance:Newsfunding_round
AI drug developer Isomorphic replenishes its coffers with $5B
Despite turbulence in the broader AI industry, investors have handed Isomorphic the two biggest cash injections in the biotech sector this year. Source: google.com
Nvidia's interim CEO: Chip shipments to China have stopped
Nvidia is no longer shipping its Hopper data center chips to China. This comes even though the US government approved the sale of a version of that model to the Chinese market. Source: bloomberg.com
Importance:PolicyAI chip export controls
Huang scandal drives bipartisan push to bar AI chip sales to China
Members of Congress want several chip-related export control bills attached to the annual National Defense Authorization Act. The effort is gaining support across party lines. Source: cnbc.com
Importance:NewsAI chip policy impact
Trump's proposed AI chip rules unsettle Southeast Asia
The Trump administration's plan to limit China's access to advanced AI chips abroad sent shockwaves through Southeast Asia on Tuesday. The proposal puts planned data center projects in the region at risk. Source: nytimes.com
Importance:Newsdata centers
Datacenters' Future Hinges on Generating Power Behind the Meter
Datacenters are no longer just passive electricity consumers. They are turning into active players in power generation, storage and grid management. Source: theregister.com
Published by Martin Ševčík
18 November 2026 at 06:11
Five billion dollars is a strange number to hand a company that, by its own admission, hasn't yet put a drug on the market. Yet that is what investors just did for Isomorphic Labs, the drug discovery startup founded by Demis Hassabis, with reports putting the valuation around $50 billion. Notice the timing, too: this Series C landed only months after Hassabis's departure from Google. I think that matters more than the headline figure. It suggests the market no longer believes the most ambitious AI science work has to live inside a hyperscaler's walls, and that a founder with a credible track record can raise sovereign-fund money for a thesis rather than a product.
My scepticism here is practical. Drug discovery is a business where a brilliant model still has to survive years of clinical trials, and biology has a long habit of humbling clever software. Still, if Isomorphic can shorten even one stage of that pipeline, the second-order effect is enormous: pharma companies will have to decide whether to partner, buy or build, and that decision will ripple through the whole industry.
Meanwhile, Google is doing what Google does, shipping at a relentless pace. Between the Gemini 3.5 Flash and Spark agent announcements at I/O and the more recent Gemini 3.7 Flash and Pixel 11 roundup, the cadence is almost dizzying. Most of this is incremental, and I would say so plainly: faster, cheaper models and tighter device integration are steady progress, not a leap. But the personal agent push deserves attention, because agents are where the real competition is heading. By the way, that same agentic capability is exactly what worries me in the third story of the day.
The FBI's report on ARTEX, an open-source agentic AI platform being used by cyber actors, and Google's parallel warning that threat actors are treating AI tools as accelerators, should land harder than it probably will. Open-source agents are a gift to builders, and I use that word sincerely. But the same properties that make them useful, autonomy and low cost, make them attractive for attackers who need to move faster than defenders can patch. If you are building with agents, the practical lesson is to treat security as a design constraint from day one, not a feature for version two.
On a calmer note, two items point to where the underlying technology may go. Q.ANT is demonstrating a photonic NPU working alongside an AMD GPU to run a language model, with Fourier transforms handled in optics rather than electronics. If that scales, it addresses the energy problem that quietly shadows every funding round above. And MIT's finding that LLM internal representations mirror physical laws is intriguing, though I would want to see how well it generalises before reading too much into it.
So here is the question I keep turning over: if capital is flooding toward AI that discovers things, and the same tools are accelerating those who break things, which side of that race is currently better funded?