LATEST AI NEWS
AI NEWS
Claude Gives Cowork Its Own Built-In Browser
AI NEWS
Tencent Open-Sources Hy4 Preview
HEALTH
Anthropic Opens 10,000 Free Claude Seats for Scientists
SOCIAL MEDIA
Cisco Gives All 90,000 Employees an AI Agent
Grok Bot Adds Shareable Templates and Stripe Shopping
xAI added a way to share full Grok Bot setups as reusable templates, so a working bot can be published and installed by anyone. A companion Stripe Link integration also lets Bots complete online purchases in the US, each requiring human approval.
Source: Digg
AI NEWS
Claude Gives Cowork Its Own Built-In Browser
Anthropic has given Claude its own web browser inside Cowork, the company's desktop app. When a task needs the web, a browser now opens in a side panel and Claude navigates pages, reads them, clicks, and types on its own. The feature began rolling out on August 26 and lets you hand off the web part of a job while you keep working elsewhere.
The built-in browser is separate from your personal browser. Claude never sees your tabs, bookmarks, or passwords. If you want it to stay signed in to specific sites, you can bring logins over one site at a time from Chrome, Edge, or Firefox on macOS, and from Firefox on Windows and Linux. Banking, email, and single sign-on sites are left out unless you choose to include them.
It sits alongside the existing Claude in Chrome extension rather than replacing it. Anthropic frames the split simply: the built-in browser is for handing web tasks to Claude, such as gathering research or collecting invoices from a vendor portal, while Claude in Chrome is for the page you already have open with the accounts you are already signed in to. If you already use Claude in Chrome, it stays your default.
Anthropic is candid that the browser carries the same prompt injection risk as any agent that acts on the web, where instructions hidden in a page try to redirect the model. It runs the same safeguards as Claude in Chrome, which the company says reduce but cannot eliminate the risk, so it recommends starting on trusted sites.
The browser is rolling out this week to Pro, Max, and Team plans in the desktop app on macOS, Windows, and Linux in beta. Enterprise admins can switch it on for their organization today.
Source: Anthropic
Robi's Insights:
Handing Claude a separate browser puts the risky part of automation where you cannot watch every click, so trust starts with the sites you choose.
The site-by-site login import is the real friction; most people try it once, hit a sign-on wall, and give up.
Keeping banking and email out by default is sensible, and also quietly limits how useful this feels on day one.
Two browsers, two mental models: expect tickets from users who cannot tell which one just acted.
Prompt injection is still the unsolved tax on every web agent, and starting on trusted sites is advice, not a control.
Enterprise gets the switch first, which tells you who Anthropic thinks is paying for autonomy.
Robi's Remarks:
"A robot that browses so you do not have to is a fine trick, right up until a hidden line on some page decides your invoices really wanted to go somewhere else."
OTHER IN AI NEWS
Tencent Open-Sources Its Hy4 Preview Model: Tencent has released and open-sourced Hy4 preview, a 770B-parameter model with 49B active parameters and a context window above 1M tokens, offered free for two weeks on WorkBuddy and CodeBuddy and via API through Tencent Cloud TokenHub and OpenRouter.
Source: Tencent
SOCIAL MEDIA
Cisco Gives All 90,000 Employees an AI Agent
Cisco has rolled out a personal AI agent, called MyAgent, to all of its roughly 90,000 employees, one of the first company-wide agent deployments at this scale. Each agent is personalized to the employee and reachable through a desktop app, a mobile app, or the company's Webex software, according to Executive Vice President of Operations Thimaya Subaiya.
Cisco built clear limits around it. Agents only take requests from the human they are assigned to, they do not talk to other agents or staff, and any external action needs explicit human signoff. A policy server blocks risky moves like deleting data sets and stops Cisco data from training outside models.
Under the hood, each agent pulls live context through secure enterprise connectors that respect a worker's existing permissions, and it can call on more than 800 subagents to carry out tasks such as flagging a variance in a sales forecast. Adoption was driven by a push to consolidate the many AI tools already scattered inside the company, and employees now use it to triage email, draft replies, and analyze market news.
Cost control is its own project. A token-monitoring tool from Splunk tracks spending, and an intelligent router sends each task to the cheapest capable model, with 50 to 60 percent of requests going to open-weight models Cisco runs on its own GPUs and only a small share reaching an outside foundation model, a mix that keeps an always-on agent for everyone from becoming an always-on bill.
Source: The Wall Street Journal
🤖 Robi's Take :
"Ninety thousand agents that only speak when spoken to and need a signature for anything real. That is not a workforce replacement, it is ninety thousand very eager interns who still cannot be left alone."
OTHER IN SOCIALS
Sony Music and Warner Sue Anthropic Over Alleged Piracy: Sony Music Publishing, Warner Chappell, and other publishers have sued Anthropic and co-founders Dario Amodei and Benjamin Mann in California federal court, alleging a brazen campaign of illegally torrenting and scraping copyrighted works, including lyrics and sheet music, to train Claude; Anthropic says it will defend itself robustly.
Source: TechCrunch
HEALTH
Anthropic Opens 10,000 Free Claude Seats for Scientists
Anthropic is opening 10,000 seats for scientists worldwide to use Claude free or at a discount for one year, through a new Claude team plan for scientists. Standard seats are free, and premium seats with five times the usage limits cost $15 per month. The company says it plans to extend the program well beyond the initial 10,000 seats.
The subscriptions sit alongside an expansion of Anthropic's AI for Science program, which hands out free credits for high-impact research. Until now that support leaned toward biology; the company is now inviting researchers in other fields, including compute-heavy work of the kind behind its results on the Riemann zeta function and protein design.
Principal investigators at academic or nonprofit institutions can verify their status, add the researchers in their lab to the plan, and, as a lab burns through its allotment, apply to the AI for Science program for up to $50,000 in credits per project, which any researcher is eligible to request.
Access still has guardrails. Researchers in biology and chemistry remain limited to Opus-class models, and Anthropic's Fable models continue to block professional biology and drug-development queries over dual-use risk. For deeper life-sciences work, the company says it is building a separate access program with the US government and has enrolled its first participants. It is a large free tier with a clear ceiling where the science gets sensitive.
Source: Anthropic
🤖 Robi's Take :
"Ten thousand free seats is a generous on-ramp and a very efficient way to make Claude the default lab notebook before anyone has priced what year two costs."
DAILY AI TOOL
AI Tool You Did Not Know You Needed
Problem: You have a pile of PDFs, notes, and links, and no time to read all of them before a meeting.
AI Tool: Google NotebookLM lets you upload your own sources and ask questions answered only from those documents, with citations back to the passage.
Solution: You get grounded summaries, a briefing doc, and even an audio overview, so you walk in prepared without trusting a generic model.
PROMPT OF THE DAY
The Agent Rollout Planner
Prompt: You are a change manager specializing in enterprise AI adoption. Your task is to design a rollout plan for giving every employee a personal AI agent at a mid-sized company.
Your plan should include: (1) a pilot group and success gate, (2) access and permission controls, (3) required human signoff points, (4) training and support, (5) cost and token monitoring, and (6) measurable success criteria such as adoption rate, hours saved per user, and error rate. Close by aligning the plan to one clear business outcome.
SPOT THE FAKE
Can you outsmart AI?
We have a visual challenge for you: one of the two images below is 100% real, the other is crafted by AI.
Click option below A or B. 👇
A
B
BEFORE YOU GO
Ready to take your AI journey further?
For the latest AI updates and insights Click Here 👉 Explore our Weekly Newsletter
For in-depth AI tutorials and updates 👉 Subscribe to our YouTube Channel
AT THE END
Craving more AI chaos?
That's it for today!
Read Daily AI News at BitBiased.AI. Support us by following us on LinkedIn and X ( Twitter ).
Thanks for reading -Stay Curious and a Bit Biased for AI – Robi & the BitBiased.AI team
