SAP Patches Four CVSS 10 Flaws as Anthropic Prices IPO at $1.4 Trillion

AI & Machine Learning

Anthropic set a price range of $70 to $78 per share for its IPO in an updated filing, offering 825 million Class A shares to raise between $57.75 billion and $64.35 billion. At the top of the range the implied fully diluted valuation is $1.404 trillion, up from $965 billion in May. Seven strategic partners, including Google, Nvidia, Microsoft, AMD, Broadcom, Micron and TSMC, agreed to a separate $26 billion private purchase of Class N stock, with Google and Nvidia taking $10 billion each. The filing puts Anthropic’s run-rate revenue at $81 billion as of late October, with Nasdaq trading under the ticker “ANT” expected around November 19. Source: TechCrunch Verified: True

Meituan published the open weights of its LongCat-2.5 vision-capable language model on Hugging Face under a permissive MIT license, a week after first introducing it. The mixture-of-experts model activates about 48 billion of its roughly 1.6 trillion parameters per token. Meituan reports a jump to 88.4 on Terminal Bench v2.1 from 70.8 for its predecessor, and to 64.6 from 9.7 on DeepSWE v1.1, while its hosted API is temporarily priced at $0.30 input and $1.20 output per million tokens. The release adds another strong Chinese open-weight coding model to a field where US labs are increasingly competing on price. Source: VentureBeat Verified: True

Stanford’s One Hundred Year Study on AI published its 2026 AI100 panel report, the third in its five-year series, working through 12 standing questions about AI’s impact on society. In an unusually candid twelfth takeaway, the panel concedes the report will already be out of date by the time anyone reads it, given the pace of model releases. It also describes global AI governance as fragmented and discloses that AI tools were used for transcription, summarization and draft checks in producing the report itself. The admission highlights how hard long-horizon academic assessment has become when frontier capabilities shift month to month. Source: The Register Verified: True

Consumer Hardware

Meta brought its Muse AI agent to the Apple Watch and Android’s Wear OS, letting users talk to it through voice calls, text messages or dictation directly from the wrist. The launch follows the Muse iPad app a month ago, the new Spaces feature on November 6 and US video chat the week before. Muse hit 10 million US downloads within its first 40 days, a pace only a dozen apps have matched. Expanding to smartwatches puts Meta’s assistant on the same devices where Siri and Gemini are deeply integrated, raising the stakes in the race to own the always-on personal agent. Source: TechCrunch Verified: True

Reddit rolled out “Reddit Reviews” to all US users on iOS, Android and the web, using AI to turn community discussions into product review pages under a “what people are saying” heading with a positive, negative or mixed breakdown. The feature starts with about 10,000 products after testing since February, with Home & Garden and Fashion & Apparel categories added in September. It positions Reddit as a direct rival to traditional review sites and shopping search, where “reddit” is already a common query suffix. The obvious risk is astroturfing, and Reddit is leaning on its existing rules and anti-spam systems to keep brands from gaming the summaries. Source: TechCrunch Verified: True

Adobe used its MAX conference in Miami to release a local MCP server that exposes Illustrator, Photoshop, After Effects and Premiere features to third-party chatbots including ChatGPT, Claude and Gemini. Firefly gains a private-beta prototyping tool called Builder that takes aim at Figma, while Photoshop, Premiere and Lightroom each get new AI-assisted features. Adobe’s Topaz Labs unit is also releasing two new models, Wonder 4 and ImageHDR. Opening its apps to rival assistants is a notable strategic shift, betting that creatives will increasingly start work inside general-purpose chatbots rather than Adobe’s own interfaces. Source: TechCrunch Verified: True

Cybersecurity

SAP’s November 2026 security updates fix 35 vulnerabilities, including four rated a maximum CVSS 10 across NetWeaver, Web Dispatcher and Gateway. They include two memory-corruption bugs in the Internet Communication Manager and Web Dispatcher that researchers at Onapsis dubbed “ICM2SHELL,” a missing-authentication flaw in the SAML2 service provider, and a broken access control bug in Gateway called “PROXYGATE” that enables pre-authentication remote code execution. Onapsis reported three of the four issues. With CISA having flagged 14 SAP flaws as exploited since 2021, administrators of business-critical SAP landscapes are being urged to patch quickly. Source: BleepingComputer Verified: True

Barracuda warned that attackers are exploiting two Email Security Gateway flaws patched just a week earlier, CVE-2026-116606 and CVE-2026-111560, to gain root access. The first is an authentication weakness that allows changes to global configuration, and the second is an unauthenticated OS command injection that runs as root. Exploitation requires access to the admin interface and has so far hit a small number of organizations. The bugs also affect Barracuda’s Load Balancer ADC, Message Archiver and WAF Control Center, making exposed management interfaces an urgent priority to lock down. Source: BleepingComputer Verified: True

Google shipped an emergency Chrome update for CVE-2026-125350, a high-severity use-after-free in the browser’s XML component that it says is being exploited in the wild. Fixed versions are 157.0.8103.50 and .51 on Windows and macOS and 157.0.8103.50 on Linux. It is the eighth Chrome zero-day exploited in attacks this year. Google is withholding technical details until most users have updated, the standard practice for actively exploited browser bugs. Source: BleepingComputer Verified: True

ESET documented “Cereal,” a new modular .NET toolset the Russian state-backed Turla group is using to spy on government Microsoft Exchange servers. It was first found at a Kazakh government body in late 2025 and then at a Romanian agency compromised in December 2025, where it selectively harvested emails exchanged with NATO and EU representatives. Components include CerealBox, which injects into Exchange’s EdgeTransport process to hijack transport agent handlers, and CerealNeuron, which listens for encrypted web requests. ESET saw no external command-and-control traffic, underscoring how stealthy on-server email implants can be, and has published indicators of compromise. Source: BleepingComputer Verified: True

Vanderbilt University is restricting access to most of its systems after an active cyberattack claimed by a ransomware group calling itself BYOD, which says it stole data including more than a million Social Security numbers. The university announced the incident on November 9 and has not confirmed the gang’s claims. Campus Wi-Fi was restored that evening and classes are continuing as scheduled. BYOD has also claimed an unconfirmed attack on Texas Health and Human Services, suggesting an active new crew targeting large public institutions. Source: The Register Verified: True

Enterprise Infrastructure

Anthropic launched Claude Meetings in beta for Team and Enterprise customers, placing Claude inside video calls as a visible participant that drafts plans as people talk, maps timelines on a whiteboard and answers questions in chat. It works with Google Meet at launch and automatically leaves if an external guest joins unless an admin allows it. Anthropic also introduced Claude Whiteboards and Claude Sheets in beta, made Claude Data generally available, and added interactive connectors for Adobe, Atlassian and HubSpot. The bundle pushes Claude further into workplace collaboration territory dominated by Microsoft 365 and Google Workspace. Source: TechCrunch Verified: True

Policy & Regulation

Nvidia now faces a securities class action after its stock fell 4.27% on Monday when its board opened an investigation into CEO Jensen Huang’s relationship with an executive US intelligence believes spied for China. Rosen Law Firm filed the suit in the Northern District of California on behalf of a retail investor, naming Nvidia and Huang and covering purchasers from October 9, 2025 to November 8, 2026. At least four other law firms announced investigations the same day. The litigation moves the affair from a national-security story into a corporate-governance fight over what Nvidia disclosed to shareholders. Source: TechCrunch Verified: True

EU agency eu-LISA has taken IBM, Atos and Leonardo to court in Estonia over the development of the bloc’s Entry/Exit System, the biometric border-control platform whose contract grew from €142 million to €212 million before it went live in October 2025. The civil claim was filed in December 2025 and confirmed by Estonia’s Harju County Court, though no hearings are scheduled yet. The EU anti-fraud office OLAF has also found sufficient grounds for further investigation. The case is a rare instance of an EU body suing its own large IT contractors and could shape how future public-sector technology deals are policed. Source: The Register Verified: True