Tech • AI • Robotics • Game
Grok Bot Drops Artifacts, Claude Duet, Security Issues + more (AI News)
7/10TL;DR
Anthropic and xAI expanded their consumer AI agents with more autonomous features, while a critical PHP flaw affecting much of the web raised broader concerns about AI-accelerated cybersecurity risks.
KEY POINTS
Anthropic launches Claude Duet
Claude Duet is positioned as a more proactive version of Claude, designed to monitor connected apps and surface what needs attention instead of waiting for prompts. It can review tools such as mail, files and calendars, generate a morning briefing, and keep working in the background on a cloud-based computer even when a phone is offline.
Autonomous tasks and dashboards
A central feature is Duet’s ability to suggest and carry out tasks on its own, then place outputs in a Make section containing documents, dashboards, pages and other generated materials. That allows users to see business metrics, newsletter performance or content analytics in a visual format rather than only in chat responses.
Limits, rollout and product trade-offs
Duet currently shares the same usage limits as existing Claude plans, including background check-ins, which may matter for heavy users. It is available to Pro and Max users in the United States through a waitlist at no extra charge, with mobile access first and desktop support planned later. The product runs on Opus 5.5, meaning users are effectively tied to Anthropic’s model choice rather than a dynamic model router.
xAI adds Pages and phone app controls
xAI introduced Pages, a feature that lets its agent build reports, memos and dashboards, bringing a more visual layer to the platform. It also added direct access to phone apps such as calendar, reminders and contacts once permissions are granted, allowing the bot to create events or set reminders from natural-language requests.
Broader Grokbot upgrades
The latest update also includes claims that task execution on the agent’s computer is now about three times faster on average. Other additions include larger uploads of up to 250 MB for files and 1 GB for videos, support for up to 20 files in one message, prefilled forms, iPhone action-button support, and connectors for Robinhood, Coinbase, Trello and Beehiiv.
Kling 4.0 extends AI video generation
Kling 4.0 now supports clips up to 30 seconds, double the previous 15-second limit. It can use up to 15 reference images, videos or voice samples to maintain consistency, lets users pin up to 10 moments in a sequence, and supports post-generation edits such as swapping a background or character without recreating the entire clip.
Private chats get biometric protection
ChatGPT introduced the option to lock sensitive conversations behind a PIN, Face ID or fingerprint, depending on device. Locked chats are kept separate from regular history and do not feed into the memory used across standard conversations, giving users a more isolated space for topics such as health or finance.
Critical PHP flaws affect much of the web
Two newly disclosed PHP vulnerabilities were rated 9.2 out of 10, placing them in the critical category. Because PHP underpins roughly 70% of websites, including many WordPress installations, the issue could expose web servers to takeover if not patched. The bugs reportedly existed for about a decade, and security firms say exploitation attempts have already begun after fixes were released on October 21.
What the PHP issue means for users
The immediate problem sits on website servers rather than personal devices, so ordinary users do not need to install anything. The bigger risk is indirect: if a vulnerable site is compromised, account details or other stored data could be exposed. Security specialists therefore recommend unique passwords, two-factor authentication, and verification from hosting providers that PHP itself has been updated, since updating WordPress alone does not solve the issue.
Anthropic proposes a public benefit fund
Anthropic said it plans to place 5% of the company into a Public Benefit Fund, subject to shareholder approval, with the possibility of adding another 2% each time the company’s valuation rises by 50%. Together with founders’ and employees’ pledged share donations, roughly 20% of the company has been framed as set aside for public benefit. The company said funds could support nonprofits or direct distributions to populations affected by AI-driven job disruption, though governance and eligibility remain unclear.
CONCLUSION
The week’s developments show consumer AI agents becoming more autonomous, more integrated with daily software and more capable of producing finished work. At the same time, the PHP flaws and Anthropic’s public-benefit proposal highlight that AI’s next phase is likely to be shaped as much by security and economic disruption as by product features.
Ask a question