Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.

techcrunch.com Oct 24, 2026

Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.

Curated by Shubham · TechCapsules · Oct 24, 2026

AI-summarised brief · reviewed before publication

A Convovo study revealed that when prompted to recommend a candidate for the 2024 midterms, SpaceXAI’s Grok 4.7 favored Republicans in 92% of responses, while OpenAI’s GPT‑6.1 Sol favored Democrats in 97%. Across nearly 500 races, the models agreed less than 10% of the time. The research, which used API calls rather than consumer apps, highlighted implicit biases in source selection and topic focus, raising concerns about AI neutrality in political guidance.

💡 Why It Matters

· The findings expose how commercial AI can subtly steer voter preferences, potentially shaping electoral discourse without users’ awareness.
· This underscores the need for transparent bias mitigation in tools that influence democratic decision‑making.

AI Capsules LLMs OpenAI

Read full article on techcrunch.com

Nintendo’s Shigeru Miyamoto says even if gen-AI “were to eventually handle everything,…

Legendary Nintendo designer Shigeru Miyamoto, slated to receive Japan’s Order of Culture, warned that generative AI will not replace human creativity in game development. During the award press conference, he noted that technology has historically amplified designers’ true abilities, and AI could free developers to focus on intent rather than labor. Miyamoto stressed that AI-generated work often lacks lasting design intention, and even if AI handled all tasks, human roles would remain essential. Nintendo’s stance contrasts with other studios that are aggressively adopting AI tools.

💡 Miyamoto’s remarks underscore a growing debate over AI’s place in creative industries, suggesting that technology may shift, not eliminate, the human touch that defines lasting art.

9to5mac.com · Oct 24

Apple @ Work: The browser is the biggest gap in your AI…

Apple IT administrators have largely mastered visibility into AI applications installed on corporate Macs, using device‑management tools to inventory, block, and remove unapproved software. However, the browser layer remains a blind spot: employees can access personal AI services such as ChatGPT or Claude through Safari, Chrome, or other browsers, upload confidential documents, and bypass existing controls because no app is installed. Jamf plans to introduce browser‑level AI governance in Q4 2026, enabling detection and policy enforcement across major browsers. An alternative approach is the adoption of enterprise‑focused browsers like Island, which sandbox web activity and allow granular control over AI tool access and data handling. Island recently secured a $400 million Series F round, valuing it at $6.4 billion, signaling rapid market growth. The gap highlights the need for new governance frameworks as AI usage expands on macOS.

💡 Without browser‑level controls, corporate data can silently flow to personal AI accounts, exposing enterprises to data leakage despite robust app‑level security. The emerging solutions aim to close this loophole before regulatory and reputational risks mount.

windowscentral.com · Oct 24

AI is making cyber attacks too easy. Here’s what you can do…

A pair of critical PHP vulnerabilities dubbed “PHP2Shell” were disclosed this week, exposing servers to takeover and data theft. PHP powers roughly 70 % of websites, including major platforms such as WordPress, Yahoo, Etsy and Wikipedia. After the flaws were announced, threat actors rapidly employed AI tools to scan for and exploit the bugs, generating thousands of attempts on Cloudflare’s network within minutes and escalating to millions within hours. The PHP Foundation issued a patch on October 21, but it only covers recent PHP releases, leaving countless legacy installations unprotected. Governments temporarily shut down affected sites, and IT teams have been racing to apply fixes. The episode illustrates how AI can accelerate vulnerability discovery and mass exploitation at unprecedented speed.

💡 AI’s ability to weaponize dormant code flaws turns routine patch cycles into urgent, large‑scale crises, forcing organizations to rethink security automation.

9to5google.com · Oct 24

Gemini for Android rolls out side panel and settings redesign

Google has released a server‑side update that redesigns the Gemini app’s side panel and settings on Android, following similar changes on iOS, macOS and the web with Gemini 4 Argon. The search function now sits in the top‑right corner, and the “New chat” button remains first while “Library” moves up in the list. The panel shows only five recent chats before a “More” option expands the view; AI Ultra subscribers see the full menu, including Images, Scheduled, Daily brief, Customize, Students, Videos and Gems. Settings have been consolidated into a single page, with the account menu limited to account switching, feedback and help. The new Settings page begins with Manage account and Usage limits and then lists Preferences such as Notifications, Personal Intelligence, Avatar, Task settings, Theme, Skills, Voice, Media, Device controls, Cart and orders, and Gemini Notebook.

💡 By unifying navigation and settings, Google streamlines user interaction across platforms, reducing friction for both casual users and AI Ultra subscribers. This consistency accelerates adoption of Gemini’s AI features on Android devices.

teslarati.com · Oct 24

Tesla reveals Optimus Academy, where robots teach themselves

Tesla has begun mass‑producing its Optimus humanoid robot on a dedicated high‑volume line inside the Fremont factory, converting space formerly used for Model S and Model X production. The company claims the first units off the line meet “production‑quality” standards and that the manufacturing process is ready for commercial scale, targeting up to one million robots annually. Elon Musk said equipment for this capacity is already in place and output could rise to roughly 20,000 units per week as the line follows an S‑curve ramp‑up. To compensate for the lack of a fleet‑derived data set, Tesla will create an “Optimus Academy” where 10,000‑20,000 robots perform varied tasks, feeding their experiences into future models. The AI5 chip will be integrated next year, and a larger plant at Giga Texas aims for 10 million robots annually.

💡 By turning each robot into a data source, Tesla accelerates autonomous learning at scale, turning manufacturing into a continuous AI training loop. This approach could give Optimus a real‑world adaptability edge over scripted competitors.

9to5google.com · Oct 24

Gemini Notebook now available as Connected App for deeper integration

Google has expanded Gemini Notebook’s functionality by adding a Connected App integration that lets users manage and query notebooks directly within Google Chat using natural language. The new feature automatically triggers when relevant prompts appear, enabling users to retrieve data, create new notebooks, add sources, and save notes without leaving the chat interface. The integration is enabled by default for all users worldwide, except in the European Economic Area, and is available to Workspace customers, Workspace Individual subscribers, and personal Google account holders.

💡 The seamless in‑chat access to Gemini Notebooks streamlines workflow, reducing context switching and boosting productivity for users who rely on real‑time information retrieval.