siliconainews.com

Yesterday · Thu 19 Nov

Google splits Gemini history from training controls

Google splits Gemini history from training and keeps chats two months

AI-generated summary Google said on November 16, 2026 that it is separating Gemini chat history from generative AI model training. Google stated that users will be able to keep conversations in their chat history without using them to train generative AI models.

Also covered byRR Tech PulseTechCrunchSilicon AI News
Earlier:

Wednesday 18 November

Introducing GPT-6.2 Astra and Sol

AI-generated summary OpenAI says it has introduced GPT-6.2 Astra and GPT-6.2 Sol as new models designed for professional work and everyday tasks. OpenAI says GPT-6.2 Astra delivers improved performance on complex professional assignments while GPT-6.2 Sol offers efficient capabilities for everyday use.

Also covered byRSS.bz9to5MacSilicon AI News
Earlier:

Put Gemini to work for you

AI-generated summary Google says an update for Gemini rolling out for hundreds of millions of people in the U.S., Japan, Korea, Canada, Australia, and India uses the Gemini 4 Hydrogen model. Google states that Gemini connects to emails, calendars, and payment methods to handle complex tasks such as scheduling appointments, managing bills, and shopping for groceries.

Also covered byData Studios ‧ExafinSundar Pichai (@sundarpichai)Silicon AI News
Earlier:

Tuesday 17 November

DAILY AI BRIEF 🗞 — Nov 17 MODEL FORECAST 👀: - MAI-Code-2 and MAI-Code-2-Flash, today!

AI-generated summary MAI-Code-2 and MAI-Code-2-Flash appeared in GitHub Copilot's model catalog with 1.05M context according to reports. Microsoft updated MAI Playground with a redesign ahead of the Ignite keynote where new models are expected.

Also covered byUnite.AIRR Tech PulseSilicon AI News

Monday 16 November

FBI Artex warning: AI pentest tool breaches in 3 hours

FBI Warns Criminals Are Using Artex to Go From Scan to Breach in Under Three Hours

AI-generated summary The FBI warned U.S. organisations that criminals are using Artex, an open source AI penetration-testing framework, to compromise financial firms, governments and law offices in under three hours. Microsoft says Storm-0940 used the framework to attempt sign-ins to more than 2,250 ConnectWise tenants, while Barracuda reported attackers used it to reach cloud accounts, an analytics system and Kubernetes clusters.

Also covered byInccrypto.newsSilicon AI News
Earlier:

OpenAI will embed Apollo researchers to investigate AI subversion

AI-generated summary OpenAI will embed six safety evaluators from Apollo Research into its organization as part of a pilot program. The pilot will investigate whether models attempt to subvert their own or their successors' alignment or control mechanisms.

Also covered byUnite.AIBPDATA AI NewsSilicon AI News
Earlier:

Hyperscale data centers’ energy consumption, carbon emissions newly estimated by researchers

Aerial view of a large data center campus in Hutto, Texas, with three massive buildings with 400M square feet of space and 600MW power capacity

AI-generated summary A study led by Harvard T.H. Chan School of Public Health and the Harvard Data Science Initiative provides facility-level estimates of electricity consumption and carbon emissions for hyperscale data centers supporting artificial intelligence. The research estimates that U.S. hyperscale data centers used between 68 and 99 terawatt-hours of electricity and were associated with about 25.7 million metric tons of carbon dioxide during the study period.

Also covered byLatitude MediaCASRAISilicon AI News
Earlier:

US Issues Rule Blocking China's Remote Access to Advanced AI Chips Abroad

US Issues Rule Blocking China's Remote Access to Advanced AI Chips Abroad

AI-generated summary The US Commerce Department issued an interim final rule to restrict remote access to advanced American artificial intelligence chips and computing infrastructure in third countries. The regulation extends export licensing requirements for Nvidia chips to seven nations and requires buyers to disclose end users and intended applications.

Also covered byThe News92TechCrunchSilicon AI News
Earlier:

Google Project Zero will disclose bugs after 90 days, fixed or not

AI-generated summary Google Project Zero reports that large language models help attackers find and exploit bugs faster while cutting the time needed to turn a patch into a working exploit. Google Project Zero will disclose vulnerabilities 90 days after reporting them to a vendor, reducing the deadline to 60 days on February 2, 2027, and to 30 days on January 3, 2028.

Also covered bySecurityWeekSC MediaSilicon AI News
Earlier:

Sunday 15 November

Apple launches research study to explore the use of AI for health

AI-generated summary Apple has launched the Apple Health AI Usage Study in the United States in collaboration with OpenEvidence to evaluate trust, comprehension, and real-world health AI engagement. Apple says the virtual research study and associated app are designed to advance the understanding of how people interact with conversational health AI tools.

Also covered byTechstreet Nowaiunderstanding.orgSilicon AI News

Saturday 14 November

Gemini 4 Hydrogen is Google's new workhorse for AI agents, and what European teams should check first

AI-generated summary Google launched Gemini 4 Hydrogen on 12 November 2026 to replace Gemini 3.8 Flash in the Gemini app, Google Search and the Gemini API. Artificial Analysis reports that Gemini 4 Hydrogen scores 51 on the Intelligence Index and costs $1.50 per million input tokens.

Also covered bySilicon AI NewsWritebleLotook
Earlier:

OpenAI’s new Daybreak agents test apps and draft fixes

OpenAI’s new Daybreak agents test apps and draft fixes

AI-generated summary OpenAI has introduced new Daybreak security agents called Red Team and Defense Factory to test applications and draft fixes for vulnerabilities. OpenAI says the agents operate within isolated environments and utilize the GPT-6 Cyber model for security tasks.

Also covered byTokenstromSI Agent News
Earlier:

Runway's open-source Praxis-1 turns video AI into a robot controller

AI-generated summary AI company Runway has released its robot model Praxis-1 with open weights and source code under the Apache 2.0 license. The model uses camera images, text instructions, and position information to predict environmental changes and control robots.

Also covered byNakoda NewsSilicon AI Newsbyteiota | From Bits to Bytes
Earlier:

How does ChatGPT Wallet work? Face ID, then a payment token · AI Agents Info

AI Agents Info — The biggest directory of AI tools and agents

AI-generated summary OpenAI says ChatGPT Wallet is a new feature allowing users in the United States to complete purchases through dots after verifying with Face ID or Touch ID. OpenAI says the payment network provides a substitute payment token so actual card details are not shared with the merchant or the agent during checkout.

Also covered bySierraPPC LandSilicon AI News