Apollo

Google Deepmind is on a roll lately Compelte Google AI ecosystem visualized view.caduceusapp.eu/apollonv… (i) Genie 3 -world model (ii) Gemini 4 Carbon/Argon -frontier llms (iii) Gemini omni 1.2 flash - text/image to video (iv) WeatherNext 3 - probabilistic generative weather forecasting model (v) Gemini Robotics 2.5 -AI for robotics their model availabiltity though is not so great

Microsoft Could Pay Gamers To Train AI (sneaky microsoft) Microsoft has published a patent application describing a system that could pay gamers for using their idle GPUs to train AI models. The application was filed on 30 April 2025 and published on 5 November 2026 under number US 20260349820 A1. The proposed system would divide large AI training workloads into smaller tasks distributed across gaming PCs and laptops and smartphones. Device owners would need to explicitly authorize the use of their hardware before participating. The software would monitor GPU utilization and memory consumption and temperature and power draw to identify available computing capacity. Training tasks would automatically pause or stop when users need their hardware or temperature limits are reached. Results from participating devices would be validated and combined to update AI models while accounting for unreliable or delayed computations. Microsoft describes potential compensation through monetary payments and software credits and access to trained AI models. The application's 20 published claims do not require compensation and an additional 250W of power consumption for four hours daily would consume approximately 30 kWh per month. Microsoft has not announced a commercial service or supported hardware list or payment rates or launch date.

ESET Identifies Russian Malware Inside Government Exchange Servers ESET published research on 10 November 2026 identifying a previously undocumented malware toolkit called Cereal targeting Microsoft Exchange servers. Researchers first detected the toolkit at a government organization in Kazakhstan on 28 November 2025. A second compromise was identified at a government organization in Romania on 23 December 2025. The Romanian infection was configured to collect emails exchanged with NATO and European Union institutions. The CerealBox component intercepts and filters email messages by operating inside the legitimate Microsoft Exchange transport process. Another component called CerealNeuron provides a passive backdoor that receives commands through incoming network requests. The toolkit does not require a separate external command server because the compromised Exchange server performs that function. Its components use encrypted payloads and execute malicious code in memory through specialized .NET modules. ESET attributed the operation with high confidence to Turla which is a cyberespionage group linked to Russia. The researchers published technical analysis and indicators of compromise to help organizations detect the malware on affected systems.

France Proposes 18 Rules For AI Assistants France's Council on AI and Digital Technology submitted 18 recommendations for regulating consumer AI assistants to President Emmanuel Macron on 10 November 2026. The proposals were developed by nine independent experts following 55 consultations with researchers and industry representatives and other stakeholders. The report covers conversational AI systems including ChatGPT and Gemini and Grok and other digital assistants. It recommends restricting features designed to make AI appear excessively human or repeatedly flatter users. Providers should clearly identify AI interactions rather than allowing users to mistake generated responses for human communication. The council also proposes mandatory identification of commercially influenced responses and a prohibition on undisclosed advertising influence. Users should be able to inspect personal characteristics and preferences inferred by AI systems from their conversations. Consumer AI systems would undergo assessments of potential user vulnerabilities before entering the market. Additional protections would address minors and emotional dependency alongside ongoing monitoring of deployed assistants. These measures remain recommendations rather than binding law and several would require changes to European regulatory frameworks.

Infineon Targets €20 Billion Revenue In 2027 German semiconductor manufacturer Infineon forecast fiscal 2027 revenue of approximately €20 billion on 10 November 2026. The forecast represents expected revenue growth of more than 20% compared with fiscal 2026. Nearly €4 billion of projected revenue would come from power supply components used in data centers. Infineon manufactures semiconductors that regulate and convert electricity for computing infrastructure and industrial equipment. The company reported fiscal 2026 revenue of €16.353 billion. Annual revenue increased 12% compared with the previous fiscal year. Fourth quarter revenue reached €4.706 billion while segment operating profit totaled €1.119 billion. The quarterly segment profit margin reached 23.8%. Adjusted free cash flow totaled €1.925 billion during fiscal 2026. Infineon expects approximately €2.2 billion in adjusted free cash flow during fiscal 2027 and plans to increase its dividend from €0.35 to €0.38 per share.

Barracuda Email Security Gateway Faces Authentication Bypass Barracuda confirmed on 9 November 2026 that attackers were exploiting two vulnerabilities affecting its network and email security products. The vulnerabilities were first disclosed through a security advisory published on 2 November. CVE-2026-116606 is an authorization bypass vulnerability with a CVSS severity score of 9.3. CVE-2026-111560 is a command injection vulnerability rated 8.6. The first vulnerability affects Barracuda Email Security Gateway while the second affects several products including web security and load balancing systems. Attackers exploiting the authorization bypass do not require valid credentials if they can reach the affected administrative interface. Combining the two vulnerabilities could allow unauthorized access and execution of commands with root privileges. Barracuda confirmed observing exploitation attempts affecting a limited number of organizations. The company has released security patches for affected products and recommends upgrading to the latest firmware versions. Barracuda also recommends restricting administrative interfaces to trusted internal networks to reduce exposure to these attacks.

Google Patches Eighth Exploited Chrome Zero Day Of 2026 Google released a Chrome security update on 9 November 2026 addressing nine vulnerabilities including one already exploited in attacks. The actively exploited flaw is identified as CVE-2026-125350. It is a memory safety vulnerability involving the browser's XML processing component. The vulnerability is classified as high severity and involves a use after free condition. An anonymous security researcher reported the flaw to Google on 6 November. Google confirmed that exploitation exists in the wild but did not disclose details about the attacks. This is the eighth actively exploited Chrome zero day vulnerability patched during 2026. The update also fixes six additional high severity vulnerabilities affecting components including V8 and WebGL and GPU functionality. Two medium severity vulnerabilities involving uninitialized GPU resources were also corrected. The fixed versions are Chrome 157.0.8103.50 and 157.0.8103.51 for Windows and macOS and Chrome 157.0.8103.50 for Linux.

SAP Web Dispatcher Receives Emergency Security Fix SAP released 35 new security notes and four GitHub advisories during its November 2026 security patch cycle. The company also updated three previously published security notes. Four vulnerabilities received the maximum CVSS severity score of 10.0. The first involves memory corruption in the Internet Communication Manager component of SAP NetWeaver. The second affects SAP Web Dispatcher through a similar memory corruption vulnerability. Researchers at Onapsis reported that these two vulnerabilities can allow remote attackers to execute operating system commands without authentication. A third critical vulnerability involves broken access controls in SAP Gateway. The fourth concerns missing authentication checks in the SAML2 service provider component of SAP NetWeaver and ABAP Platform. Additional vulnerabilities received severity scores of 9.9 and affect components including NetWeaver Knowledge Management and the ABAP Application Server. SAP released security fixes on 10 November and recommends that affected organizations prioritize installing the updates.

Microsoft Opposes US Ban On Chinese Data Center Equipment Microsoft is urging the Trump administration to reconsider a proposed ban on Chinese optical transceivers used in American AI data centers. Reuters reported the discussions on 10 November 2026 citing three people familiar with the matter. Optical transceivers enable high speed communication between servers and switches through fiber optic connections. Chinese manufacturers account for approximately 66% of the global market according to Counterpoint Research. Microsoft currently has 35 data centers under construction in the United States. Another 134 Microsoft data centers are planned according to infrastructure tracker Cleanview. The proposed restrictions are being considered by the Federal Communications Commission over concerns about possible espionage and network interference. Microsoft and other technology companies have warned that a broad ban could delay construction because American suppliers cannot immediately replace Chinese production capacity. US officials are reportedly considering a narrower restriction targeting transceivers capable of transmitting 3.2 Tbps. Production of those components is not expected to expand significantly before 2028 and no final prohibition has been announced.

Harvey Adds Danish Legal Sources Harvey announced a partnership with Danish legal publisher Djøf Forlag on November 11 2026. The deal adds licensed Danish legislation executive orders and preparatory works to the Harvey legal AI platform. Jurabibliotek licensees gain a way to research and draft from those materials alongside their own documents. Jurabibliotek legislative content is scheduled to become available in Harvey at the start of 2027 with no additional charge. Access requires a license to Djøf Forlag digital platform. The licensed collection will sit alongside Danish public sources already available in Harvey. Djøf Forlag Digital Director Bjarke Madsen said the publisher has supplied Danish legal sources for more than 65 years. The content can be used in Harvey Assistant Agents Vault Workflows and the Word Add-in.

Armada Opens Dublin Data Center Plant Armada announced on November 10 2026 that continuous production had started at Galleon Forge Two in Dublin. The site is a 90,000-square-foot manufacturing operation inside a Johnson Controls facility. The first Dublin-built Leviathan modular data center was completed for Fossefall in Norway. The site is Armada's first European manufacturing operation and the second Galleon Forge built with Johnson Controls. Fossefall's initial order covers five Leviathan systems totaling more than 9 MW with delivery scheduled by the first quarter of 2027. WinDC plans to deploy 11 MW of modular data-center capacity across renewable-energy sites in Australia. Approximately 700 Johnson Controls engineers and technicians support the Dublin operation. The companies are targeting at least 500 MW of capacity in 2027 as a forward-looking goal.

Anthropic Prepares Stock Market Listing Anthropic was founded five years ago by a breakaway group from OpenAI. Dario Amodei joined OpenAI in 2016 and ran its safety team and scaling laws research. Daniela Amodei joined OpenAI in 2018 and became vice president of safety and policy. The Amodeis and several colleagues left OpenAI in December 2020 and launched Anthropic four weeks later. Sam Bankman-Fried invested $500 million through FTX in nonvoting shares. That FTX stake was sold for $884 million to repay the exchange's customers. In May Anthropic raised $65 billion at a $965 billion valuation. In February Trump ordered federal agencies to stop using Anthropic's technology.

Mac Studio M5 Ultra Takes On DGX Spark Apple's Mac Studio with the M5 Ultra SoC ships with a 36 core CPU and an 80 core GPU plus up to 512GB of unified memory. The tested configuration costs $12,300 and current shipping waits run 13 to 15 weeks. The M5 Ultra offers 1.2 TB/s of memory bandwidth versus 273 GB/s on Nvidia's GB10. On the AgentPerf-Local benchmark the Mac Studio finished in just under half the time of a single DGX Spark class system. The Mac Studio also consumed about 16% less task energy in that test. A single Dell Pro Precision GB10 with a 4TB SSD now costs $10K. A pair of DGX Sparks costs $15K or more which is 25% more money than one M5 Ultra Studio. With the TensorFold engine GLM 5.3 Flash

Infineon Sees No Slowdown in AI Data Center Buildout Infineon said Tuesday that the rapid buildout of AI data centers would not slow down. The German chip maker reported a 12% rise in revenue for its fiscal year ending in September. Infineon forecast sales growth climbing to 20% in the current year due to AI data center demand. CEO Jochen Hanebeck acknowledged growing debates about expanding AI infrastructure. He called the resistance a rather American-centric topic. Hanebeck said operators and American hyperscalers will find areas outside the United States for new sites. Infineon does not make AI processors itself and supplies chips that deliver electricity from the grid to the sites. Hanebeck said AI delivers real value and its importance will only grow.

Mercor raises $750M at $20B valuation Mercor.io Corp. raised $750 million in a Series D round at a $20 billion valuation. The company draws on a network of more than 5 million domain experts to train and evaluate AI models. Annualized revenue passed $2 billion in June according to co-Chief Executive Brendan Foody. The round drew corporate money from Nvidia Corp. Salesforce Inc. and Zoom Communications Inc. Returning investors Benchmark Capital Management Co. Felicis Ventures Management Co. and General Catalyst Group Management also participated. The prior round was a $350 million Series C in October 2025 that valued Mercor at $10 billion. Mercor bought New York startup Deeptune Inc. in July and part of its post-training runs on Deeptune software. The new money will pay for research and evaluation infrastructure including more enterprise application environments for AI agents.

FMF Releases Jailbreak Rating Framework The Frontier Model Forum released a Jailbreak Rating Framework to grade how easily AI models escape safeguards. The project started in 2023 with Anthropic Google Microsoft and OpenAI as founding companies. The framework is a rough analog to the CVSS security rating system but targeted at LLM jailbreaks across multiple frontier providers. Version 1.0 is a work in progress and the group accepts comments until December 21. The framework has four scoring components split across severity and likelihood categories. Each component is graded from zero to three and the combined rating runs from low to critical. A capability gain of zero means the jailbreak offers no improvement beyond existing models or tooling. A capability scope of three means the jailbreak works across at least two unrelated categories of cyber warfare.