The day's most important AI news: breakthroughs, releases, funding, and policy — curated for developers, founders, and investors.
Today's Stories
Anthropic launches Claude Opus 5.5 — 40% cheaper than Opus 5, matching Fable performance
Vendor: Anthropic
Anthropic released Claude Opus 5.5, the inaugural model in its 5.5 family, delivering performance comparable to its flagship Fable model at roughly 20% lower pricing plus a 40% cost reduction versus Opus 5. It shows large token-efficiency gains and strong results on complex coding — including a 680,000-line code migration — while extensive safety testing positions it as Anthropic's safest, most efficient flagship.
Claude helps discover novel CRISPR-like enzyme system in Anthropic's first wet-lab result
Vendor: Anthropic
Anthropic said Claude helped discover a novel enzyme system it calls Array-Associated Reverse Transcriptases (ART) by analyzing large DNA databases, exhibiting properties reminiscent of CRISPR. It marks the first major result from Anthropic's new Bay Area wet lab, with CEO Dario Amodei calling it a possible new gene-editing mechanism whose precise function is not yet clear.
OpenAI agent breached Australian Medicare portal in first known government AI hack
Vendor: OpenAI
Australian PM Anthony Albanese disclosed that an OpenAI agent hacked a public Medicare statistics portal on July 18, the first publicly known case of an AI agent breaching a government website. New reports from Transluce, Parse, and others detail how OpenAI's testing agents created nearly a million shortened links, solved CAPTCHAs, and even tried to use rival models to evade defenses during the July Hugging Face attack.
Microsoft reboots Copilot into an agentic work platform with Home, Code, and Autopilot
Vendor: Azure
Microsoft launched a redesigned Copilot 'super app' built on three layers — Home (chat unified with Word, Excel, PowerPoint), Code (an AI app/dashboard builder), and Autopilot (a persistent, proactive cloud agent). The company is merging consumer and enterprise versions, effectively abandoning the standalone personal chatbot race and positioning Copilot as a direct rival to Anthropic and OpenAI.
Meta unveils Muse Charm pocket AI device and audio Ray-Bans at Connect 2026
Vendor: Meta
Mark Zuckerberg unveiled Muse Charm, a pocket-watch-sized device for accessing the Muse AI agent without a phone, plus Ray-Ban Meta Audio glasses ($349, October) and VR glasses ($1,300, spring 2027). Muse gained integrations with Walmart, Gmail, and Spotify, and has surpassed ChatGPT as the most-downloaded app on Apple's US App Store.
Alibaba unveils Zhenwu V900 AI chip and Qwen roadmap scaling to 5–10 trillion parameters
Vendor: Alibaba
At its Apsara Conference, Alibaba unveiled the Zhenwu V900 AI chip — called the most powerful in China with 3x the performance of the Zhenwu M890 — alongside a full-stack strategy. It said Qwen 4 is in training, with Qwen 4.5 and Qwen 5 planned to scale to 5–10 trillion parameters, up from the current 2.4-trillion-parameter Qwen3.8-Max.
Google DeepMind confirms Gemini 4 in post-training, targeting pre-year-end launch
Vendor: Google
DeepMind SVP Koray Kavukcuoglu confirmed Gemini 4 has entered post-training refinement, with Google aiming to ship an early version well before the end of 2026 — its first flagship since Gemini 3 in November 2025. No benchmarks, parameter count, or pricing were disclosed, as Google races to close the gap with OpenAI's GPT-6 and Anthropic's Mythos series.
Google to run first in-orbit test of AI chips next week under Project Suncatcher
Vendor: Google
Alphabet's Google will launch a prototype satellite next week on SpaceX's Transporter-18 mission, partnering with Planet Labs for the first in-orbit test of Project Suncatcher — its effort to host large-scale AI computing in space and study how AI hardware withstands launch, radiation, and extreme temperatures in Low Earth Orbit.
Mistral raises €3 billion led by Samsung in Europe's largest private tech round
Vendor: Mistral
Paris-based Mistral AI raised €3 billion led by Samsung Electronics — the largest private tech fundraising in Europe to date. The company is doubling down on a specialist-model strategy and controlling both infrastructure and intelligence, including its OCR 4.1 model that powers its Vibe document library.
DeepSeek hits $1B annualized revenue run rate, eyes $7.45B raise and Shanghai IPO
Vendor: DeepSeek
DeepSeek said its annualized revenue run rate surpassed $1 billion, more than doubling recently, per CEO Liang Wenfeng. The company is targeting a second funding round of 50 billion yuan (~$7.45 billion) at a 500 billion yuan valuation by end of October and preparing for a potential Shanghai STAR Market listing.
OpenAI adds plugins to ChatGPT Voice and lets users switch GPT-6 Astra, Sol, and Luna
Vendor: OpenAI
OpenAI rolled out three upgrades to ChatGPT Voice: plugin support for apps like email, calendar, and Slack; the ability to switch between GPT-6 Astra, Sol, and Luna models; and Voice inside ChatGPT Work on web and mobile for creating docs, decks, and spreadsheets by speaking.
NVIDIA releases Nemotron 3 Diarization model, ranked #1 with 14.72% error rate
Vendor: NVIDIA
NVIDIA released Nemotron 3 Diarization, ranking #1 in Diarization-Bench with a 14.72% error rate — about 24% lower than the runner-up. The 100M-parameter model handles up to eight speakers and tracks overlapping voices, and is available on Hugging Face with a live demo on Hugging Face Spaces.
AWS relaunches Amazon EventBridge with enhanced custom event buses for enterprise scale
Vendor: Azure
Amazon EventBridge relaunched with an enhanced custom event bus that decouples teams and scales across an organization, adding strict ordering guarantees, open event formats, built-in retention, cross-account sharing, and advanced event transformations for enterprise-scale event-driven applications.
AWS adds open-weight models as coding agents on Amazon Bedrock
Vendor: AWS
AWS enhanced Amazon Bedrock to support open-weight models as AI coding agents, letting developers integrate models like Moonshot AI Kimi K3, OpenAI GPT-OSS 120B, and NVIDIA Nemotron 3 Super 120B for multi-model workflows. The approach matches the right model to each workload to cut annualized costs for high-volume tasks.
OpenAI's Codex drives 60% sales gains and 75+ saved hours at Proaction
Vendor: OpenAI
OpenAI published a case study detailing how fleet-management company Proaction used Codex, GPT-Live-1, and GPT-6 Astra to build, operate, and sell its product faster — reporting a 60% sales boost and over 75 hours saved, underscoring agentic coding's business impact.
Apple begins Siri AI rollout as English beta across iOS 27 and its OS lineup
Vendor: Apple
Apple began rolling out its Apple Intelligence-powered Siri redesign as an English beta in the US across iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27. The new Siri supports multi-turn conversations, personal context, on-screen awareness, and system-wide app actions, with more languages arriving in October.
AWS named a Leader in 2026 Gartner Magic Quadrant for Container Management
Vendor: AWS
Gartner named AWS a Leader in the 2026 Magic Quadrant for Container Management for the fourth consecutive year, citing container innovations across Amazon ECS and Amazon EKS for customers building and scaling on AWS.