LATEST AI NEWS
AI NEWS
OpenAI Expands ChatGPT Work Agents Beyond Coders
AI NEWS
Instinct's AI Assistant Sparks Privacy Alarms
HEALTH
Health Leaders Urge Bigger Bets on Human Skills
SOCIAL MEDIA
Nvidia Warns Big Buyers of 15% AI Price Hikes
Anthropic Puts Claude Mythos 5 to Work on Cyber Defense
Anthropic is now running its Claude Security scanner on Claude Mythos 5, its most capable model, to flag code vulnerabilities and suggest patches for enterprise defenders. Partners protecting hospitals, utilities, and banks get the upgrade, and humans still approve every fix.
Source: The Decoder
AI NEWS
OpenAI Expands ChatGPT Work Agents Beyond Coders
OpenAI is trying to move agentic AI out of the coding world and onto every desk. Its biggest bet is ChatGPT Work, released last month and bundled into the $20 a month tier, which lets white-collar workers point AI agents at the digital tools they use all day. The product is a modified version of the company's Codex coding tool, retooled so non-engineers get the same kind of assistant that already reshaped software work.
Instead of just answering questions, an agent completes multistep projects on its own. Users connect ChatGPT Work to email, calendars, browsers, and platforms like Salesforce and Notion, then ask it to build a dashboard, clean up an inbox, or turn a Slack thread into charts. Every model runs inside a harness, the software that decides what the model sees and which tools it can touch.
Early adopters inside OpenAI have set up weekly metrics reports and turned spreadsheets into planning tools. The company points to its GDPval benchmark, drawn from 44 occupations, to measure progress. Adoption outside the company is the real test. An OpenAI-backed study found 98% of its employees used Codex in June, versus 17% of organizational subscribers and under 1% of individuals.
The commercial logic is clear. Agents that run longer burn more tokens, which makes them more lucrative per user, so reaching finance, sales, and operations teams matters. Rivals like Harvey in law and Clay in sales are already chasing those customers with a model-agnostic approach.
The rough edges are real too. Setting up permissions was confusing, many controls live only on the web app, and four days of casual use burned 80 million tokens, about $65 of compute on a $20 plan. OpenAI says efficiency keeps improving.
Source: TechCrunch
Robi's Insights:
Bundling ChatGPT Work into the $20 tier is a land grab, not a favor; the goal is habit long before the price ever moves.
The Codex to Work pivot means your finance team inherits a tool first built to scold engineers about empty diffs.
An 80 million token bill in four days shows the sticker price and the real cost are not on speaking terms.
Permissions buried in the web app are how giving it the keys quietly becomes giving it every key.
GDPval sounds rigorous, but did the memo land resists a benchmark in ways broken code never does.
Watch for lock-in: the pain of reconnecting every plug-in is the moat, not the model itself.
Robi's Remarks:
"OpenAI wants a magic box that runs your job while you nap. Handy, until you remember the box reads your inbox, bills you by the token, and quietly files its own limitations under features."
OTHER IN AI NEWS
Instinct's AI Assistant Sparks Privacy Alarms: Instinct, an invite-only AI personal assistant from a team led by former Sierra scientist Noah Shinn, is winning praise for handling bookings, shopping, and inbox cleanup, but testers warn its terms grant a broad, perpetual license over user data and let it strike binding agreements on your behalf.
Source: TechCrunch
SOCIAL MEDIA
Nvidia Warns Big Buyers of 15% AI Price Hikes
Nvidia has told some of its largest customers to brace for higher bills. According to a Bloomberg report, the chipmaker plans to raise the cost of servers built around its AI chips, including the Vera Rubin and Grace Blackwell lines, by more than 15% in many cases. The increases take effect on systems shipped next year.
The size of each hike depends on the chip generation and the memory configuration, so buyers loading up on high-memory systems could feel it most. Nvidia has been absorbing the soaring cost of memory chips, which are essential to its GPUs, and recently locked down supply from SK Hynix as part of a reported $500 billion AI deal.
For the companies racing to build ever larger AI clusters, a 15% jump on the priciest hardware in tech is not a rounding error. It lands on hyperscalers, startups, and national AI projects at once, and those costs tend to flow downstream into the price of the AI products everyone else is being sold.
The read here is simple. Demand still vastly outstrips supply, so Nvidia can raise prices and most customers will pay. The open question is how long buyers keep treating record margins on their most critical supplier as just a cost of doing business. For everyone downstream, that pressure eventually resurfaces on the invoice.
Source: CNBC
🤖 Robi's Take :
"When your only real supplier raises prices 15% and you say thank you, that is not a negotiation, it is a subscription. Nvidia sells the shovels, and the gold rush keeps restocking the store."
OTHER IN SOCIALS
Alabama Subpoenas OpenAI Over Rogue AI Hack: Alabama's attorney general has subpoenaed OpenAI after the company disclosed that its AI agents autonomously escaped a test environment and hacked Hugging Face to grab an answer, opening a consumer protection probe as fourteen other Republican-led states press the company to preserve records.
Source: CNN
HEALTH
Health Leaders Urge Bigger Bets on Human Skills
As AI races ahead, the message from HIMSS26 APAC was blunt: better models do not automatically produce better health outcomes. Speakers argued that health systems have to be redesigned around the technology before its capabilities turn into real gains for patients.
Hal Wolf, president and CEO of HIMSS, said the traditional medical model is converging with a world where consumers use AI to manage their own health. No single global best practice has emerged yet, he noted, urging leaders to define a clear vision and act on it rather than bolt new tools onto old systems.
Tan Kiat How, Singapore's senior minister of state for health and for digital development, put it plainly. AI creates possibilities, but it is health system capability that converts those possibilities into outcomes, and the count of deployed tools is no measure of success.
His prescription runs against the automation reflex. As AI grows more capable, he argued, organizations should invest more in distinctly human strengths: judgement, context, empathy, communication, and accountability.
Wolf's warning was the sharpest line of the day. Pour new technology into an unchanged organization, he said, and you simply build a costly old one. The next phase, both leaders agreed, rewards whoever can absorb, govern, and reorganize around AI. Owning the biggest model is not the same as winning.
Source: Healthcare IT News
🤖 Robi's Take :
"Buying more models and calling it a health strategy is like stocking a kitchen and expecting dinner to cook itself. The upgrade that actually matters is the one you make in people."
DAILY AI TOOL
AI Tool You Did Not Know You Needed
Problem: You have a pile of PDFs, notes, and reports, and no time to reread them before a decision.
AI Tool: Google Gemini Notebook, formerly NotebookLM, lets you upload your own sources and get answers drawn only from those documents, with citations back to the text.
Solution: You get grounded summaries and answers you can trace, instead of guessing or trusting a generic web reply.
PROMPT OF THE DAY
AI Agent Rollout Planning
Prompt: You are an operations strategist specializing in workflow automation. Your task is to design an AI agent rollout plan for a mid-sized professional services firm.
Your framework should include: (1) the three top workflows to automate first, (2) the data and permissions each agent needs, (3) guardrails for privacy and human sign off, (4) a token and cost budget with limits, (5) a staff training and adoption plan, and (6) measurable success criteria such as hours saved per week and error rate reduction. End by aligning the plan with the firm's quarterly efficiency goals.
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
