Wednesday, November 18, 2026

Top 5 This Week

Related Posts

Gemini 4 Carbon: Google advances its cybersecurity results | Brief AI


Google announces 100% scan recall for Carbon in an internal evaluation of its Fortify system. The model becomes accessible via API to paying customers, while its arrival to Google AI Ultra subscribers continues. Its announced uses range from scientific research to coordinated teams of agents in Antigravity.

Google claims flaws discovered in SaaS providers

In Scan for Good, Wiz’s program dedicated to public digital infrastructure, Google reports that Carbon identified 0-day vulnerabilities at two large SaaS providers in less than two hours. The company indicates that this operation made it possible to reduce work that would have taken several weeks to one afternoon. Wiz worked with affected organizations to remediate the vulnerabilities in record time, limiting the potential impact on tens of thousands of businesses.

The results of Fortify, Google’s autonomous system covering scanning, sorting and remediation, come from an internal evaluation: Google announces 100% recall for scanning and 95% precision for sorting. On open source repositories, the company claims a relative improvement of more than 25% in recall compared to Argon, with the detection of previously unidentified logic faults. Google says it is among the largest users of open source software and sees this progress as benefiting both its own systems and the broader ecosystem.

Another illustration: Google’s CVR team, experts in cloud vulnerabilities, used Carbon to prove the exploitability of two bugs present in widely used open source software. Google rates the severity of these flaws at a level comparable to that of the PHP2Shell exploit and indicates that a proof of concept carried out with Carbon led the maintainers to reclassify the bug as critical and release a patch without delay. Technical details will be communicated after the responsible disclosure procedure has been completed.

Google describes encouraging practical feedback, both internal and external, for this version which it presents as the most advanced in cybersecurity. Via Fairwind, it claims strong performance on SEC-bench Pro for detection, ExploitGym for exploitation and CWE-bench v1 for vulnerability correction. Software inspection, cross-system defect chain analysis, and source-verified patch generation are among the announced capabilities.

Axiom Bio transfers its analyzes to Carbon and Argon

Axiom Bio, a company specializing in drug toxicity, considers Carbon to be the most successful model in translational science and credits it with the discovery of mechanisms of toxicity that it had not previously identified. The transfer of its scientific production analyzes and its agents to Carbon and Argon is underway, with the aim of designing safer medicines.

Google claims to improve Carbon’s reasoning in physics, mathematics and life sciences, to interpret experimental data, test hypotheses and solve multi-step problems. Its researchers used Carbon in Antigravity for enzymatic assays, with only laboratory notebooks and raw data as input. According to Google, the model highlighted difficult-to-interpret anomalies and diagnosed experimental artifacts, including incomplete automated mixing in deep wells, which helped improve reproducibility. Google also mentions better methods for deconvulsing chromatographic peaks and detecting trace samples.

For software verification, researchers designed AlphaProof Code and applied it to Signal’s cryptographic protocol code. Carbon, combined with other models, was used to translate security objectives into mathematical theorems; the agents then generated millions of lines with it in Lean, a mathematical proof language. Google reports that key security properties were proven in a matter of weeks, whereas this work historically took specialized teams years. The company sees potential for extending formal verification on a larger scale, without generalizing this result to all software.

Google charges Carbon at Argon’s introductory rate

API access to Carbon is available through Google AI Studio and Gemini for Business, for all paying developers and businesses. The launch price is identical to that of Argon: $2 per million incoming tokens and $10 per million outgoing tokens. Rolling out to all Google AI Ultra subscribers is underway in Antigravity, the Gemini app, and Search AI mode.

The evening letterFree · 7 p.m.

Are you passionate about AI research?

The papers and advances that matter, explained simply, every evening. Free.

Included upon registration: our selection of the best AI guides & comparisons.

Choose your pace

Free · No spam · Unsubscribe in 1 click

Carbon is accessible to developers, businesses, scientists and consumers. It succeeds Gemini 4 Argon and was first offered, two weeks before this announcement, to a group of cybersecurity advocates via the Fairwind program. According to Google, Carbon represents its most advanced model for long and complex agentic tasks, while Argon remains preferred for the majority of uses in software engineering and knowledge management. Carbon’s use cases include sophisticated code development, debugging, or open scientific research, and it uses a larger volume of tokens than Argon to enable deeper exploration, reasoning, and verification.

In Antigravity, Google highlights the /swarm command: a single instruction allows you to launch a coordinated team of agents, whose activities are monitored live. Carbon can be assigned to architectural, auditing or coordination missions, alongside faster, lighter models operating in isolated spaces in parallel. Adversarial verifiers test each change.

Google cites multi-package migrations, root cause searches between subsystems and complex architectural redesigns among the targeted uses. For web applications, the company describes an operation alternating construction, error management and checks of the code as well as the displayed interface.

Carbon comes second on the MKA multimodal assessment

Google and Artificial Analysis launched the Multimodal Knowledge Agent (MKA) benchmark to evaluate multimodal capabilities in agentic tasks. Google considers that existing benchmarks do not correctly measure agentic reasoning on dense files mixing images, graphs, tables, plans and diagrams. On MKA, Carbon ranks second, with an average cost per task more than 6X lower than the first place model.

On CursorBench 4.0, Perplexity’s WANDR and Zapier’s AutomationBench, Google gives Carbon the first position. These evaluations are described as independent and representative of concrete tasks, with a cost per task more than half that of the next best external model. According to Google, the first financial analyzes indicate a performance equivalent to that of the best model tested at Hebbia on E2E Financial Agentic.

Carbon also comes first among 23 configurations on General Financial Intelligence from Balyasny Asset Management, according to Google. This test covers law, the use of agentic tools and their execution, programming, data engineering, and perception and contextual understanding. For legal, financial and business operations professions, Google describes a model combining documentary research, visual understanding and reasoning on large sets. He also attributes to him the production of tables, presentations and reports subject to strict formats.

/swarm demonstrations range from building to 3D gaming

Architectural demonstrations include Google Platform 37: seven public municipal planning PDFs, including 206 extracted technical views, enabled interactive reconstruction. Google announces 906K triangles in Three.js, after a setpoint and 10 nightly loops with /swarm. Another team of agents created, also from a single instruction, an interactive 3D exhibition on the construction of the Eiffel Tower between 1887 and 1889, using historical photographs and engineering boards.

In the scientific field, Google presents a multiscalar textile visualizer obtained in Antigravity from academic work, with few instructions. It models virtual fibers by finite elements, with 3D wire meshes, polar tracking of 110 fibers and dynamic stress-strain curves. Another technical achievement: a Stein-Schule fortepiano from 1780, built from CAD plans and x-rays, playable on 61 keys with escapement physics. The instrument is embedded in a 3D representation of Koerner Hall at the Royal Conservatory of Music in Toronto, with seat-dependent spatial acoustics.

Google also credits Carbon with visual and spatial advancements applied to gaming and creative code. An open world game features a pelican on a GBike on the Mountain View campus: its sub-agents used Blender with Omni and Nano Banana references, Three.js and WebGL shaders, then Web Audio for procedural sound generation and Lyria for musical composition. An underwater sequence of Atlantis begins with a video produced with Gemini Omni; its procedurally generated ruins, volumetric caustic lighting, and water simulation in Blender are then exported to Three.js for real-time rendering in the browser.

On the commerce side, Google reports a redesign of the Google Merchandise Store obtained in two instructions, with five artistic directions and product visuals generated by Nano Banana 2.1. This achievement mobilizes front-end design and motion design skills in Antigravity. Online versions of the store, game and Fortepiano are offered, as well as a Gemini front-end design skill.

Shredder Smith
Shredder Smith
Shredder Smith is the lead curator and digital persona behind topaitools4you.com, an AI directory dedicated to "shredding" through industry hype to identify high-utility software for everyday users. Smith positions himself as a blunt, no-nonsense reviewer who vets thousands of emerging applications to filter out overpriced "wrappers" in favor of tools that offer genuine ROI and practical productivity. The site serves as a watchdog for the AI gold rush, providing categorized rankings and transparent reviews designed to help small businesses and creators navigate the crowded tech landscape without wasting money on low-value tools.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Popular Articles