Google Research

Google Research has open-sourced RRSI, a framework allowing AI agents to refine their own evaluation harness while guarding against overfitting. Announced via MarkTechPost, the release lets developers inspect and build on the underlying code. The announcement is drawing attention from AI practitioners interested in agent self-improvement methods that remain reliable rather than gaming their own benchmarks.

The European Commission has ordered Google to share anonymized search data, according to a MediaPost report dated July 20, 2026. The move fits the EU's ongoing scrutiny of Google's dominance in search and its efforts to give rivals and researchers access to data. Details of the decision's scope and Google's response are not yet available.

A Norwegian science news report describes a study finding that children and teenagers aged 9 to 18 increasingly turn to AI assistants instead of Google for information and answers. Researchers say the shift is happening fast, but the effects on how young people learn, judge sources and handle misinformation are not yet known, prompting calls for closer attention from educators and parents.

Google researchers say the hacking group ShinyHunters has broadened its attacks targeting Oracle's PeopleSoft software. The group, known for large-scale data theft and extortion, is reportedly exploiting vulnerabilities to breach organizations using the enterprise platform. The findings raise concerns for companies running PeopleSoft, as security teams are urged to patch systems and monitor for intrusion attempts linked to the campaign.

Google researchers say the hacking group ShinyHunters has broadened its attacks targeting Oracle's PeopleSoft software. The group, previously known for large-scale data theft campaigns, is reportedly exploiting vulnerabilities to compromise enterprise systems. The finding raises concerns for organizations running PeopleSoft, with security teams urged to patch systems and review exposure to the campaign.

Security researchers are flagging CVE-2026-95356, a critical use-after-free flaw in the WindowDialog component of Google Chrome. Versions prior to 154.0.8037.57 are affected. The bug carries a CVSS score of 9.6 and could let a remote attacker, using social engineering and a crafted HTML page, run arbitrary code outside the browser sandbox. Users are being urged to update Chrome promptly.

AI researcher Timnit Gebru says in a Wired interview that she does not believe artificial intelligence poses an 'existential threat' to humanity. Gebru, known for her high-profile departure from Google's ethics team and criticism of the AI industry, argues the focus on hypothetical doom distracts from present-day harms. Her remarks are drawing attention within AI ethics debates.

AI researcher Timnit Gebru says in a new interview with Wired that she does not believe artificial intelligence poses an 'existential threat' to humanity. The interview, featured on Wired's Big Interview podcast, pushes back against prominent doom narratives around advanced AI. Gebru, known for her work on AI ethics and her departure from Google, argues the focus should be on real harms happening now rather than speculative future risks.

Researchers at Washington University in St. Louis have published findings pointing to gaps in the citations and claims made by Google's AI Overview feature, which generates AI-written answers atop search results. The study suggests the tool does not always reliably attribute or support its statements, adding to ongoing scrutiny of how AI search features handle sourcing and accuracy.