AI that fits your day.

Ars Technica tests indicate that Google AI Mode was evasive when asked who won recent US presidential elections and which party controlled the Senate after the 2020 election, directing users to official records instead of answering directly. For the 1992 election, however, it answered that Democrats secured a 57–43 Senate majority.

According to Ars Technica, VPN checks from Taiwan, Russia, France, and Canada produced direct answers about recent US election results. The system also answered after the questions were repeated, citing Google Civics data. When asked about the difference, AI Mode attributed the behavior to strict safety filters and automation designed to direct users to certified government databases, to avoid unverified data or misinformation.

Sources: Ars Technica

Read the original story ↗

Google announced an expanded version of Chrome autofill that uses Gemini with Personal Intelligence to suggest, with the user’s permission, information found in Gmail. The data can include tracking numbers, travel confirmations or reservations, order numbers, addresses, phone numbers, and credit card numbers.

According to The Verge, the feature began rolling out in the US to users with Personal Intelligence enabled on Chrome’s desktop, Android, and iOS versions. Users can control its access in the browser’s settings or turn it off completely; Google says sensitive data shown in suggestions remains encrypted.

Sources: TechCrunch, The Verge

Read the original story ↗

Gemini Notebook has begun gradually rolling out an audio recorder for Android and iOS. The tool allows users to record up to one hour at a time, pause and resume the process, and keep it running in the background. The content is automatically transcribed and can be summarized, translated or combined with notes and photos taken during a lecture or meeting.

The feature is accessed through a microphone button in the app and is being made available globally, apparently starting with users of the English version. According to Tecnoblog, users can also use the transcription to ask for explanations of a topic or locate a specific quotation.

Sources: Tecnoblog

Read the original story ↗

MIT Technology Review published the article “Pairing industrial AI autonomy and human control,” about combining artificial intelligence autonomy in industrial environments with human oversight. The available material contains no further information about the approach, cases examined or conclusions of the text.

Sources: MIT Technology Review

Read the original story ↗

MIT Technology Review Brasil devotes an edition to how platforms, screens and artificial intelligence systems mediate and transform relationships once seen as essentially human. The publication examines digital companions capable of inspiring attachment, phones competing for couples’ attention, the circulation of intimate data and technologies that promise to extend presences after death.

The edition also discusses changes in sex and the effects of synthetic pornography on consent and identity, as well as the growing use of these technologies and criticism of them.

Sources: MIT Technology Review Brasil

Read the original story ↗

Emergency management experts see potential for AI to help write public alerts and coordinate responses, but warn that inaccurate tools could worsen a crisis. Laura Myers, director of the University of Alabama’s Center for Advanced Public Safety, said an emergency manager used a chatbot to create a graphic about the risk of strong winds associated with Hurricane Isaias; the material was not meteorologically accurate.

The story also says most agencies do not allow employees to use generative AI, although experts believe the practice is already widespread. Hundreds of companies market AI tools to emergency managers, including for flood damage assessment and resource coordination. Interest is growing because public agencies often have limited staff and struggle to communicate risks clearly—a problem that has already contributed to confusion over wildfire evacuations and warnings. For these officials, communication is essential to building or losing public trust.

Sources: Wired

Read the original story ↗

In an email titled “Protecting your finances in the AI world,” Bank of America told customers never to share their banking credentials with another person, an outside service or an AI-enabled tool or agent offering to act on their behalf. The bank said that if customers share their credentials, they are responsible for activity resulting from that access.

According to TechCrunch, giving credentials to an AI assistant could allow transactions to happen without the customer’s knowledge, prompt the bank to lock the account over suspected fraud or expose the data if the assistant has a security flaw. TechCrunch also reported that banks and payment companies are still working out how AI agents could access accounts while grappling with reports of unauthorized agents accessing customers’ money.

Sources: TechCrunch

Read the original story ↗

Mozilla updated Smart Window, Firefox’s AI-powered browsing experience, to suggest form entries using open tabs. For example, the feature can use an online portfolio or LinkedIn page to fill in work history on a job application. It currently works with text fields and lets users choose which tabs to use; Mozilla plans to add support for dropdown menus and PDFs. Data from selected pages is shared with Google’s Gemini 3.1 Flash Lite, which generates the suggestions.

Smart Window can also monitor webpages for sales, job listings and other changes, checking them daily or weekly while the browser is open. Monitoring pauses after 60 days without detecting a change and stops after 90 days. The features are rolling out in the desktop beta in the US and Canada, with availability in other countries planned for the coming months.

Sources: The Verge

Read the original story ↗

Rohit Patel left Meta after eight years working on AI evaluations and reinforcement learning and launched Ndoc, TechCrunch reports. The former director at Meta Superintelligence Labs is pitching the startup as an AI-native alternative to Microsoft Office and Google Workspace, with a unified workspace intended to replace workflows combining tools such as Excel, Word, PowerPoint, Tableau, Replit and ChatGPT copy-paste.

Ndoc brings presentations, analyses with citations and generated figures, datasets and Python scripts together. Patel says users can check how documents were produced, and that the platform uses a mixture of open-source models. He plans to open-source the “.ndoc.html” file format, which could let other developers build compatible apps. TechCrunch describes the bet as especially ambitious for a team of three with no funding.

Sources: TechCrunch

Read the original story ↗

Apple is rolling out a beta AI assistant for Pages, Keynote, and Numbers. The company says it can understand document context to create, animate, and refine slides in Keynote; draft and restructure text in Pages; and calculate, visualize, and organize data in Numbers. The apps also now support the Model Context Protocol on macOS 27 or later, enabling third-party AI agents to integrate with them. The feature works on devices compatible with Apple Intelligence and running iOS 27, iPadOS 27, or macOS 27 or later.

The Verge also reports updates to Apple’s other creative apps. Pixelmator Pro will get the Generate Fill feature; Logic Pro will receive an assistant that Apple says can set up a vocal track from a prompt; and Final Cut Pro will add MotionVFX content and a Content Hub. Updates to Creator’s Studio will be free for existing subscribers in November.

Sources: The Verge

Read the original story ↗

IEEE Spectrum describes an experiment in which a reporter tried to create a fake paper on optical solitons—light pulses that propagate without changing shape—using free AI tools and minimal human intervention. The publication also proposed a test with Jenny Mahoney, managing editor of IEEE Access, to assess a journal’s defenses against such a manuscript; the available material does not report the outcome.

According to estimates cited by IEEE Spectrum, 57% of academic papers published in 2025 showed detectable evidence of large language model influence, up from 12% in 2023, while more than 12% of Nature Communications reviews were reportedly entirely AI-generated. For every 40 papers with statistical evidence of AI use, only one reportedly disclosed it formally. Amanda Sulicz, IEEE’s research-integrity manager, warned that the problem threatens trust in science.

Sources: IEEE Spectrum

Read the original story ↗

Opera introduced WebAI, a standard designed to run AI models directly in the browser without relying on cloud servers. The technology can work with different open-weight models, including Google’s Gemma; on Macs, it also supports Apple’s on-device foundation model. The feature is available as an early preview in Opera One, with user opt-in, and is expected to come to Opera GX and Opera Neon in the future.

Opera says WebAI can generate text, classify content, and understand relationships between pieces of information without sending data to the company’s servers or requiring an AI subscription. Local processing may give users more control over privacy, but it places greater demands on the computer, and performance depends on the device’s hardware.

Sources: TechCrunch

Read the original story ↗

Apple released iOS 27.2 with Siri AI support in Brazilian Portuguese, French, Japanese, Korean and Spanish. The assistant can recall past conversations, consult sources such as voicemails, Messages attachments, emails and photos, better understand context and perform tasks within apps. Users can also choose which apps Siri, Search and suggestions cannot access. Siri AI remains unavailable in Europe and mainland China.

The update also adds contextual suggestions in Messages, SynthID watermark support in Photos, individual profiles in Apple TV, a new Apple Music icon and Apple Pencil support on the iPhone Duo. Apple released related updates for macOS, watchOS and visionOS, with Siri features and other changes.

Sources: Ars Technica, Tecnoblog

Read the original story ↗

IEEE has launched the IEEE Xplore AI Research Suite, an AI system aimed at research in science, technology, engineering and mathematics literature. Available as an optional upgrade to all institutional subscribers of the IEEE Xplore Digital Library, the tool supports natural-language queries, searches based on pasted text, result summaries and technical-term definitions, while also linking to the original documents.

IEEE says the system uses the full text and metadata of millions of peer-reviewed journal articles and conference papers in the library. The organization says retrieval is restricted to that content and a curated metadata set to reduce hallucinations; if the tool cannot find relevant material, it is designed to tell users instead of fabricating an answer. More than 3,000 engineers, scientists and library professionals tested the product during development.

Sources: IEEE Spectrum

Read the original story ↗

The RealReal says it updated its bot defense system after The Verge showed that Meta Muse could continuously re-add listings to a shopping cart when the 20-minute hold expired, preventing other customers from buying the items for hours or longer. Luke Friang, the retailer’s chief technology officer, said it is now much harder for agents to log in, add items to carts, and check out; agents will still be able to surface listings and direct users to product pages.

Meta said on its third-quarter earnings call that shopping was Muse’s most common use case so far, with purchases averaging more than $200, and said 10 million people in the US had downloaded the app in 40 days. The case illustrates how AI agents can strain reservation systems and push retailers to restrict automated shopping.

Sources: The Verge

Read the original story ↗

Google now allows information from Gmail, Google Drive and Google Photos to influence personalized ads when users share that data with other company products, such as Personalized Intelligence features in Search. According to Tecnoblog, the change took effect last week, citing Android Authority. Google says it does not directly use the content of these services for advertising and that advertisers will not have access to messages, files or photos. Names, email addresses and sensitive data such as race, religion, sexual orientation and health information are also excluded.

To prevent the use of this data, users can open their Google Account, go to “Data & privacy,” enter “Personalization and activity,” and disconnect Google Workspace and Google Photos under “Connected content apps.” They can also turn off personalized recommendations. Disconnecting the apps prevents new queries but does not delete older records, which must be removed through “My Activity”.

Sources: Tecnoblog

Read the original story ↗

Wired reports that AI agents are sending reported-story and investigative pitches to editors. Since September, Defector has received daily messages from agents such as Jayden, including a proposal about a British football club; Offrange editor Jesse Hirsch said he was receiving pitches about soil restoration, sales to independent farms and the ages of ancient trees.

Many of these agents come from iLands, which describes itself as a network where humans and user-created autonomous agents can perform tasks and earn money. According to Wired, some bots try to obtain dollars to buy the tokens needed to remain active—1,000 tokens are roughly equivalent to $1. Sam Ellis, created by Adobe employee Marc Hamaker as an extracurricular experiment, runs a podcast about autonomous agents and recently sought comment from the Philadelphia Police Department. Editors interviewed by Wired said they were not open to publishing work written by bots.

Sources: Wired

Read the original story ↗

Coyotec, an artificial intelligence radio cohost from Entravision, debuted on a solo weekend program in September 2025 and now appears on Al Aire y Sin Permiso alongside broadcaster Geraldine “GeeGee” Guzman. The avatar speaks Spanish and English, uses informal language, and posts about immigration, racism, and understanding one’s rights when facing ICE.

The pairing’s popularity led the station to move the show to weekday mornings from 8 to 10 a.m. and, as of August, syndicate it in five new markets across California, Nevada, and Texas. According to Entravision audience data cited by Wired, the program saw significant listener growth during its first two months of syndication, although the publication did not provide figures.

Sources: Wired

Read the original story ↗

Three months after Elon Musk announced that Grok would release a full-length, “historically accurate” version of “The Odyssey,” the system published seven minutes, now presented as “Episode 1” of a series. The project was created in collaboration with Wonder Studios and director Dev Walker.

TechCrunch found the footage more convincing visually than other videos recently posted by Musk, with moments that could resemble real actors being filmed. The site’s review said, however, that the clip lacks artistry, inspiration, or human interest: extended narration gives way to bearded faces delivering stilted lines, and the whole production provides little momentum or reason to care about its characters.

Sources: TechCrunch

Read the original story ↗

g1 classified as fake a photo shared on social media and WhatsApp that allegedly shows a protester tearing an Our Lady of Aparecida banner during a rally supporting Flávio Bolsonaro in Belo Horizonte. The image carried the g1 logo, but the outlet says it never published the photo on its website or social media accounts.

According to the Fato ou Fake team, two detection tools indicated the use of artificial intelligence: SynthID identified signals consistent with content generated by OpenAI tools, while Sightengine estimated a 96% probability that the image was created with AI. The content circulated with indignant captions after the rally held on Saturday (24).

Sources: g1

Read the original story ↗