Google launches Personal Intelligence as Anthropic redesigns tests

Key Takeaways

• Google launched "Personal Intelligence" on January 22, 2026, for AI Pro and AI Ultra subscribers, allowing personalized search results by connecting Gmail and Google Photos.
• xAI faces EPA scrutiny over its use of unpermitted methane gas turbines at its Memphis data centers, following a new EPA rule issued January 22, 2026, clarifying permit requirements.
• Former xAI engineer Sulaiman Ghori departed after revealing details about xAI's temporary data center permits and its heavy reliance on AI agents for internal operations.
• Ukraine's Brave1 defense cluster partnered with Palantir on January 21, 2026, to create a "Dataroom" for training AI models to equip interceptor drones against Russian aerial threats.
• Carmignac's Kristofer Barrett identifies AI as the primary driver of global stock market performance, with interest expanding beyond major tech companies.
• Utahns express concerns about AI's impact on job displacement, citing past automation-related layoffs and the state's limited social safety nets.
• Salt Security released new tools on January 22, 2026, including a Databricks Connector, to enhance API security for AI and modern web systems by monitoring AI agent access.
• Anthropic's Claude Opus 4.5 models now outperform most human applicants on technical take-home tests, prompting the company to redesign evaluations and release an original test as an open challenge.
• Sid Pardeshi, a former Nvidia employee, founded Blitzy, an AI gaming company, in 2023, raising $4.4 million after leaving Nvidia in 2022.
• Hevi introduced an AI-powered "Contract Awareness Training" module to convert complex industry contracts into e-learning courses for mining, construction, and heavy industries, aiming to reduce risks.

Google Search gets personal with AI Mode

On January 22, 2026, Google launched Personal Intelligence in AI Mode for Search. This new feature allows Google AI Pro and AI Ultra subscribers to connect their Gmail and Google Photos. It provides uniquely tailored search results by using personal context from these apps. For example, it can suggest sneakers based on past purchases or create family travel itineraries using hotel bookings and photo memories. Users maintain full control, as connecting these apps is optional, and they can provide feedback to improve results.

Google AI Mode links to photos and email

Google is introducing a new feature called Personal Intelligence, rolling out on January 22, 2026. This tool allows users to connect Google's AI Mode with their Gmail and Google Photos. By doing so, the search engine can provide more personalized answers based on individual interests, habits, and travel plans. For instance, it can suggest weekend getaways based on past trips or recognize preferred clothing styles from Google Photos. Robby Stein, a Google Search VP, noted that users can correct any AI mistakes.

Google AI Mode uses personal photos and email

Google is launching a new feature called Personal Intelligence that allows its AI Mode to connect with users' Gmail and Google Photos. This option, rolling out on Thursday, aims to provide more tailored search results. For example, it can suggest weekend trips based on past travel or identify preferred clothing styles from photos. Robby Stein, a Google Search vice president, explained that this makes Search feel uniquely personal. Users can choose to turn this feature on and can correct any inaccurate AI responses.

Google AI Mode personalizes search with your data

Google is rolling out "Personal Intelligence" for its AI Mode, starting today for Google AI Pro and AI Ultra subscribers. This new Labs feature allows users to connect their Gmail and Google Photos to AI Mode. It helps the AI provide more personalized and useful answers by understanding user preferences and background. For instance, AI Mode can suggest clothing for a trip by checking flight details in Gmail. Google states that the AI models are not directly trained on personal emails or photos, and users can revoke access at any time.

xAI engineer leaves after revealing company secrets

Sulaiman Ghori, an engineer at Elon Musk's xAI, departed the company shortly after a podcast interview on "Relentless" published January 21, 2026. Ghori revealed that xAI's data center permits, including for its Colossus supercomputer in Memphis, Tennessee, were temporary and obtained quickly. He also mentioned that Colossus uses at least 35 methane gas turbines. Furthermore, Ghori shared that xAI heavily relies on AI agents for internal work, with one person managing 20 agents for API rebuilding. His departure came just days after the interview, adding to intrigue as xAI cofounder Greg Yang also left.

EPA questions xAI's use of unpermitted turbines

On January 22, 2026, the Environmental Protection Agency (EPA) issued a new rule that challenges xAI's use of unpermitted methane gas turbines. Elon Musk's AI company has been using dozens of these turbines at its Memphis, Tennessee data centers without Clean Air Act permits. xAI argued these "temporary" turbines did not require permits, a stance supported by local health officials. However, the EPA's new rule clarifies that all combustion turbines, even portable ones, are considered stationary sources and need permits. EPA Administrator Lee Zeldin denied the rule specifically targets xAI, stating the agency supports xAI's permitting process to boost the economy.

Ukraine uses Palantir AI to train drones

On January 21, 2026, Ukraine's Brave1 defense technology cluster partnered with American company Palantir to create a "Dataroom." This secure digital platform allows Ukrainian defense companies to train AI models using sensitive military data on Russian aerial threats. The main goal is to equip interceptor drones with AI to better detect, classify, and neutralize targets, especially Shahed-type drones. Ukrainian Minister of Defense Mykhailo Fedorov stated that this is crucial for countering thousands of drone threats. The Dataroom is currently only accessible to Ukrainian industry due to the sensitive nature of the information.

Thurston County Auditor keeps AI camera details secret

Thurston County Auditor Mary Hall announced she will not share specific technical details about the new AI-enabled security cameras for the Voter Registration Center in Tumwater. Hall believes revealing "blueprints" or "blind spots" of the system could help potential intruders. The Board of County Commissioners approved a $35,104 grant for these cameras, which are designed to flag threats like weapons. Hall cited the 2021 Election Security Act, HB 1068, which exempts certain election security information from public disclosure. This decision follows a public meeting where Hall accidentally shared some details and later asked for them to be deleted, which the county refused.

Carmignac says AI remains top stock market driver

Kristofer Barrett, Head of Global Equities at Carmignac, states that the artificial intelligence theme continues to be the main force behind global stock market performance. He noted that Carmignac's Investissement global equity fund significantly outperformed its benchmark in 2025. Barrett observes that the strong interest in AI is now spreading beyond large technology companies to include other industries and regions. This indicates a broader market participation in the AI trade.

Utahns worry about AI job impact

Mike Walterman, a retired computer engineer, explains that Utahns' concerns about artificial intelligence are based on real-world experiences, not just fear. Many workers in Utah have seen layoffs due to automation, leading to a distrust of how technology is used by companies. Walterman points out that Utah's weaker social safety nets for healthcare and housing make job loss particularly devastating. He argues that AI will speed up existing trends of companies reducing staff for efficiency. Therefore, Utahns are asking practical questions about what happens to them if AI leads to job displacement.

Salt Security boosts AI and edge API protection

On January 22, 2026, Salt Security launched new tools to improve API security for AI and modern web systems. The Salt Databricks Connector provides dedicated API security for Databricks environments, focusing on the "Agentic Action Layer" to see what AI agents access. This helps security teams understand and control AI workloads. Additionally, the Salt Netlify Collector offers traffic collection for Netlify's edge network, supporting modern web architectures. These new features are available now as part of the Salt Illuminate™ platform, expanding Salt's "Universal Visibility" strategy.

Nvidia employee leaves to found AI startup

Sid Pardeshi, a former Nvidia employee, left the company in 2022, just months before the major AI boom. He was inspired by Nvidia CEO Jensen Huang's advice to deeply commit to one area of expertise. Pardeshi took this to heart, focusing on AI in gaming, and decided he wanted to become a founder himself. Despite losing millions in unvested stock, he successfully launched his own AI company, Blitzy, in 2023, which has since raised $4.4 million. Pardeshi used Huang's leadership lessons to guide his new venture.

Anthropic creates AI-proof job tests

On January 21, 2026, Tristan Hume from Anthropic shared insights on designing technical evaluations that artificial intelligence models cannot easily beat. Anthropic's performance engineering team uses a take-home test to evaluate candidates, but newer Claude models, like Opus 4.5, now outperform most human applicants. Hume has redesigned the test three times to keep it ahead of the AI's capabilities. The company still needs skilled engineers and is looking for creative ways to find them. Anthropic is releasing the original take-home test as an open challenge, noting that humans with unlimited time can still achieve better results than Claude Opus 4.5.

Hevi uses AI for contract training

Hevi announced today the launch of its new "Contract Awareness Training" module, designed for mining, construction, and heavy industry. This AI-powered tool quickly transforms complex project contracts into easy-to-understand e-learning courses. Users can upload a contract, and the system generates a six-module video course with knowledge checks in minutes. Co-founder Paul Culvenor explained that this helps site teams understand contract details, reduce risks, and protect profits. The new module is available now within the Hevi platform, helping organizations ensure consistent contract awareness across all teams.

Sources

• Personal Intelligence in AI Mode in Search: Help that's uniquely yours
• Google offers users option to plug AI mode into their photos, email for more personalized answers
• Google offers users option to plug AI mode into their photos, email for more personalized answers
• Google adds your Gmail and Photos to AI Mode to enable "Personal Intelligence"
• Engineer at Elon Musk's xAI Departs After Spilling the Beans in Podcast Interview
• EPA pokes Musk over using unpermitted turbines for AI
• Ukraine feeds sensitive military data to Palantir AI for training
• Thurston Co. Auditor draws ‘hard line’ on sharing Al security camera specifics
• Carmignac's Barrett says AI trade is still the key as equity market participation broadens
• Voices: In Utah, concerns about AI are rooted in experience
• Salt Security Expands "Universal Visibility" with Specialized API Security for Databricks and Rapid Edge Support for Netlify
• Leaving Nvidia Before the AI Explosion Sounds Crazy—Until You Hear the CEO’s Lesson That Drove Him Out
• Designing AI-resistant technical evaluations
• Hevi Launches AI-Powered Contract Training to Turn Complex Contracts into Job Ready Learning in One Click