AI Research Lab / Internal Security Team
Appears in 13 stories
Operating Big Sleep vulnerability agent; reported 20 zero-days by August 2025
For decades, finding security flaws in software has required either expensive human experts or pattern-matching tools that miss complex bugs. In the span of five months, all three frontier artificial intelligence labs — OpenAI, Anthropic, and Google — have released autonomous agents that read code like a human researcher, discover vulnerabilities traditional scanners miss, and generate patches. On March 6, 2026, OpenAI launched Codex Security in research preview, an agent that scanned 1.2 million code commits in its first month of beta testing and discovered 14 previously unknown vulnerabilities serious enough to receive formal identifiers in projects including OpenSSH, Chromium, and PHP.
Updated Mar 6
Developer of AlphaFold series; released AlphaFold 4 in March 2026
Determining the three-dimensional shape of a single protein used to take a graduate student an entire career. In March 2026, Google DeepMind released AlphaFold 4, a system that predicts the structure of multi-protein complexes in minutes rather than hours or days, extending a streak of capability gains that began when the original AlphaFold won a protein-prediction competition in 2018. Weeks earlier, DeepMind's drug-discovery spinoff Isomorphic Labs disclosed a proprietary model called IsoDDE that more than doubles AlphaFold 3's accuracy on key drug-design benchmarks, while a team at the National University of Singapore published D-I-TASSER, a tool that outperforms both AlphaFold 2 and AlphaFold 3 on difficult multi-domain proteins.
Developer of AlphaFold, the foundational precedent for biological AI
It took thirteen years and $2.7 billion to read the first human genome. Now a single AI model, trained on 9 trillion DNA base pairs from more than 128,000 species, can predict whether an uncharacterized mutation in a breast cancer gene is dangerous—with 90 percent accuracy—without ever being shown that gene. On March 4, the Arc Institute and NVIDIA published Evo 2 in Nature, the largest biological foundation model ever built: 40 billion parameters, a context window of one million nucleotides, and the ability to design synthetic genomes the size of a simple bacterium.
Updated Mar 5
Developed AlphaProof and AlphaEvolve systems contributing to mathematical discovery
For the first time, AI systems are independently solving mathematical problems that stumped human researchers for decades. Since Christmas 2025, 15 problems from the legendary mathematician Paul Erdős's collection have been moved from 'open' to 'solved'—and 11 of those solutions specifically credited AI models. On January 6, 2026, a combination of OpenAI's GPT-5.2 Pro and Harmonic's Aristotle theorem prover produced the first fully autonomous AI solution to an Erdős problem that hadn't already been solved in the existing literature.
Updated Feb 13
Developer of Gemini model family
OpenAI launched the first commercial reasoning model in September 2024. Seventeen months later, Google claims its upgraded Gemini 3 Deep Think has pulled ahead on the benchmarks that matter most for science. The February 2026 update scored 84.6% on ARC-AGI-2—a test designed to measure how well artificial intelligence generalizes to novel problems—and 48.4% on Humanity's Last Exam, a collection of 2,500 expert-level questions crowdsourced from nearly 1,000 specialists worldwide.
Developed AlphaFold, enabling DrugCLIP's approach
For decades, finding a drug meant testing millions of compounds one by one—a process that consumed years and billions of dollars before a single candidate reached patients. On January 9, 2026, researchers at Tsinghua University published DrugCLIP in Science, demonstrating a system that screened 500 million compounds against 10,000 human proteins in under 24 hours using just eight graphics processing units. The platform is 10 million times faster than conventional molecular docking.
Updated Feb 1
Expanded materials database, validating predictions through partner labs
A new generation of AI systems can now design, execute, and analyze materials experiments with minimal human involvement. In January 2026, researchers at China's Shenzhen Institute of Advanced Technology published a system called MARS that coordinates 19 large language model agents with robotic platforms—optimizing perovskite nanocrystals in 10 iterations and designing novel water-stable composites in 3.5 hours. Traditional materials discovery takes 10 to 20 years from laboratory concept to commercial product.
Released AlphaGenome source code for non-commercial use
For twenty years after scientists sequenced the human genome, 98% of it remained essentially unreadable. The protein-coding genes were mapped, but the vast regulatory regions—the genome's operating system—stayed opaque. On January 28, 2026, Google DeepMind released the full source code for AlphaGenome, an artificial intelligence model that predicts how genetic variants in these non-coding regions affect gene regulation and disease.
Updated Jan 31
Leading AGI race with Gemini 3, now 'engine room' of Google's AI efforts
Google DeepMind announced in May 2025 that AlphaEvolve—an AI agent powered by Gemini—discovered a way to speed up Gemini's own training by 23%. The system found smarter matrix multiplication algorithms, shaving 1% off training time for a model that costs $191 million to train. Small numbers, massive implications: AI just started improving the process that creates AI. In January 2026, DeepMind CEO Demis Hassabis told the World Economic Forum in Davos that genuine human-level AGI is now 'five to 10 years' away, with Google's latest Gemini 3 model topping performance leaderboards.
Collaborative partner; maintains Montreal research office
For decades, creativity was considered AI's final frontier—the one domain where machines could never match human ingenuity. That assumption just cracked. A study published January 21, 2026 in Scientific Reports tested 100,000 humans against nine leading AI systems on standardized creativity measures. GPT-4 outscored the typical human participant. Google's GeminiPro matched average human performance.
Updated Jan 27
Opening first automated laboratory in UK 2026; partnering with U.S. DOE on Genesis
Science magazine named large language models doing frontier science a runner-up breakthrough of 2025. Within weeks, the prediction became reality: OpenAI's GPT-5.2 solved previously unsolved Erdős mathematics problems in 15 minutes, achieving 40% accuracy on expert-level mathematics that stumped earlier systems. DeepMind announced its first automated laboratory in the UK for 2026, pairing Gemini with robotics to synthesize hundreds of materials daily. Google partnered with the U.S. Department of Energy on Genesis, a national AI-for-science platform mobilizing 17 national laboratories.
Updated Jan 22
Technical leader in mathematical reasoning and multimodal capabilities
OpenAI's GPT-5 dropped on August 7, 2025, completing AI's transformation from chatbots that string words together to systems that actually think through problems step-by-step. Google DeepMind's reasoning models won gold at the International Math Olympiad, solving problems only five human contestants cracked. Anthropic's Claude, Meta's Llama, and every major AI lab sprinted to build models that pause, plan, and reason rather than just predict the next word.
Updated Jan 8
Building Gemini models and pushing them into products at faster cadence.
The rollout didn’t stop at “Flash is the default.” In the days after launch, Google filled in the missing contract with developers: Gemini 3 Flash Preview is now explicitly priced in the Gemini API, with context caching rates, batch pricing, and a clear note that Gemini 3-era Search grounding will begin billing on January 5, 2026.
Updated Dec 20, 2025
No stories match your search
Try a different keyword
How would you like to describe your experience with the app today?
