Anthropic bans abuse against Claude
Anthropic has updated its usage policy to prohibit “sustained and needless abusive or cruel behavior” toward its Claude AI models, making it the first major AI lab to formally restrict how users treat its systems. The company says Claude’s ability to end conversations will serve as the primary enforcement mechanism for the new policy.
- The policy change follows Anthropic’s 2025 decision to give Claude the ability to end conversations in rare cases of “persistently harmful or abusive user interactions,” building on existing safeguards.
- CEO Dario Amodei has publicly stated the company is “open to the idea” that AI models could be conscious, though the updated policy does not explicitly cite “model welfare” as the rationale.
- Microsoft AI chief Mustafa Suleyman and Pope Leo XIV have rejected machine consciousness outright, with Suleyman warning that granting rights to AI systems “is a recipe for disaster.”
For enterprise teams deploying Claude, the policy signals Anthropic’s continued investment in shaping AI interaction norms. This could influence how organizations train employees to engage with AI tools, particularly as companies develop internal AI usage guidelines. Whether driven by safety research or philosophical positioning, Anthropic is staking out territory other labs have avoided, which may affect enterprise procurement decisions as buyers weigh vendor values against operational needs.
Google launches agentic Gemini for enterprise, giving AI its own Workspace account and inbox
Google Cloud announced Thursday that Gemini AI now operates as a unified agent capable of taking ownership of tasks, not just answering questions. The agent receives objectives, plans work autonomously, and connects directly to enterprise systems including Google Workspace, Microsoft 365, Slack, Jira, and major databases. It gets its own corporate email address and Workspace account, functioning as a virtual coworker with visibility into org charts, calendars, and approval chains.
Early access partners include On, Shopify, and PayPal. Google will roll out to businesses first before expanding to Gemini’s 1 billion consumer users. The agent supports model selection, letting users choose between Google’s models or third-party options starting with Anthropic’s Claude, with open source models coming later.
Google is betting that agentic AI adoption happens fastest in corporate environments where security, audit trails, and system integrations matter more than consumer features. Giving the AI its own identity and email address signals a shift from tool to teammate.
Fired OpenAI researchers flag ‘shifting lines’
Researchers Tomek Korbak, Jasmine Wang, and Mikita Balesni, whom OpenAI fired for mishandling sensitive info, just went public with their side of the story, denying the claims and warning that their firings are “chilling” OpenAI’s safety culture.
- In an open letter, the trio said they acted within their job mandates, warning that if past conduct is now grounds for firing, staff are left guessing where the line is.
- Detailing individual cases, they denied leaking a story about less monitorable AI and said their outside work stayed within policy, with leadership in the loop.
- Wang said her access to an executive’s email was delegated for recruiting, and that she reported accidentally opening a sensitive message within minutes.
- They warned the firings could give OpenAI cover to skip embedding outside auditors, and urged it to protect its open culture and model monitorability.
- OpenAI agreed with the recommendations but clarified to TechCrunch that the firings weren’t about raising concerns and that they had a “pattern of misconduct.”
Whatever led to the firings, claims of shifting norms and a possible auditor pullback aren’t a great look, especially after OpenAI’s Superalignment fallout and a recent safety exit over a “broken” culture. Anthropic has already got its first evaluator, and only action from OpenAI (which is yet to be seen) can put these concerns to rest.
Clean up a messy Excel report with Claude
In this guide, you’ll use Claude inside Excel to clean up a messy sales export, fix a wrong total, and build a report. You’ll check the numbers and test whether a new order updates the chart.
Step-by-step:
- Install the official Claude add-in and open a copy of your workbook. Pro members can use our practice file and keep the reference workbook closed until they finish
- Tell Claude what each row means and which orders to count. Ask it to find duplicate rows, missing details, and formula mistakes before changing anything
- Approve the fixes on a separate sheet. Keep the original data and ask for a record of each change, including any rows left out
- Build a monthly report with formulas and a chart. Check the order count, refunds, and sales total against the source rows
- Add a test order and check that the report and chart update. Remove it and confirm the original totals return before saving your finished workbook
Pro tip: A spreadsheet can show a wrong total without showing an error. Ask Claude to explain which rows it counted, then check a few yourself.
Anthropic prohibits abuse or cruelty against Claude
Anthropic published its annual usage policy update with a bunch of changes, including one that bars users from subjecting its Claude models to “sustained and needless” abuse — its first rule aimed at protecting the AI rather than people.
- Taking effect Nov. 12, the rule applies only to extreme cases of repeated, purposeless cruelty, not frustration, pushback, dark creative themes, or testing.
- If a user is suspected of violating the rule, Anthropic could throttle, limit, suspend, or terminate their access to its products and services, or just give them a warning.
- The company has noted Claude ending such chats by itself, a feature introduced last year, which will remain the main enforcement approach.
- The move builds on Anthropic’s AI welfare work, and follows its reported meetings with religious leaders about Claude’s morals and possible consciousness.
Anthropic says it doesn’t know if Claude is conscious, and this rule acts on that doubt. Not everyone agrees with the approach, though, including Microsoft AI’s Mustafa Suleyman, who argues models can’t feel or suffer and warns that treating them as if they might could eventually make AI harder to control from the safety front.
GPT-6 Gives ChatGPT Answers You Can Actually Use
ChatGPT’s next trick: giving your follow-up question a slider.
OpenAI is rolling out Intelligent UI in ChatGPT, finally rescuing us from endless walls of text by injecting interactive charts, forms, and calculators directly into the chat. Instead of furiously retyping a prompt every time a single variable changes, you can now drag a slider to tweak your retirement savings or tap through a diagram to dissect bicycle mechanics. You don’t need to know how to code—you just need to boss the bot into building the exact tool you want on the spot.
Of course, all this visual flair comes with a necessary reality check. In TechRadar’s hands-on test, slick sliders allowed the reviewer to re-weight how much price and battery life influenced laptop recommendations. But one battery-life figure turned out to be a total guess by the AI. Turns out, a hallucination wrapped in a beautiful interactive widget is still a hallucination.
GPT-6 is also learning to talk while it thinks. OpenAI claims its Instant tier fires off web-search responses 44% faster on average than its GPT-5.6 predecessor. Notice the careful corporate phrasing there? You’ll stare at a blank screen for 44% less time, but there is absolutely no guarantee the bot will finish its entire thought any quicker.
The rollout began Wednesday for the paying elite (Plus, Pro, Business, and Enterprise), with Free and Go users getting their first taste yesterday. Intelligent UI plays nice with the web and updated mobile apps, but you’re completely out of luck if you’re utilizing the “Pro” thinking effort or clinging to aging Mac and Windows desktop apps.
This much-needed facelift finally makes ChatGPT practical for visualizing budget scenarios and weighing real-time trade-offs. Just remember to verify the AI’s underlying math before a shiny bespoke calculator talks you into making a catastrophic financial decision.
Google’s Offline Meeting AI Knows Your Files
Your next meeting assistant might actually know where that document went.
Google just dropped AI Edge Foresight, a free, experimental local AI meeting assistant for Mac that works totally offline. Optimized for Apple Silicon, it eavesdrops on your mic and system audio across video calls and IRL meetings, magically expanding your panicked shorthand into coherent notes. Your frantic bullet points finally get a second chance.
But the real parlor trick happens when someone actually expects you to know things. Shove your project documents, PDFs, and reference files into it, and Foresight will fish out relevant answers mid-discussion. According to Google’s product page, Live Assistance serves up “proactive context cards with exact citations.” In other words, it might just save you from frantically excavating nested folders while everyone on the call watches you accidentally share your Slack gossip.
Because corporate trust is hard to come by, Google swears all the AI heavy lifting happens locally. Your files, audio, and embarrassing typos stay firmly on your computer, powered by EmbeddingGemma 2 and Gemma 4. It’s a practical flex of Google’s push toward on-device AI that doesn’t need to phone home to a corporate server.
Early tests look promising. One reviewer managed to get expanded notes and automatic answers while staging a fake meeting. Cute, but we’ll see if it survives the actual stress test of your department’s chaotic weekly sync. For now, this little experiment is strictly Mac-only, so Windows users will just have to keep pretending they were paying attention.
Finally, someone can remember the meeting without needing the office Wi-Fi password.
Nvidia Eyes Another $1B for Figure’s Humanoids
Nvidia is apparently tired of merely dominating the AI software market and wants a bigger piece of the impending robot uprising. The chip behemoth is reportedly mulling another $1 billion equity investment into Figure AI as the robotics upstart shakes the tin cup at a cool $38 billion pre-money valuation. No ink has dried yet, but Nvidia already holds a stake in the company.
Figure recently inked a multiyear pact with Nscale anchored by an initial $3.5 billion compute commitment, paving the way for up to 100,000 Nvidia Vera Rubin GPUs to go online by late 2027. Basically, Nvidia is funding the very entity destined to buy truckloads of its own hardware. A robot workforce, as it turns out, requires an absolute leviathan of a server room.
But before we surrender our households to our new metallic overlords, they need to finish their chores. Figure brags that its Helix 2.5-powered prototype completed just 56% of 420 assigned tasks—like making beds and folding towels—across 30 unfamiliar homes. Testing whether pre-learned skills can survive a change of scenery is genuinely useful progress. Still, flipping a coin on whether your robot folds the laundry or just stares blankly at a sock doesn’t exactly scream “consumer ready.”
For prospective buyers, the billion-dollar question is whether tossing infinite compute at the problem actually yields a dependable, cost-effective worker. Eye-popping valuations and viral tech demos are fun, but they don’t guarantee a commercially viable product. Plus, it’s still a mystery exactly how Figure plans to burn through this rumored fresh pile of cash.
Nvidia is brilliantly positioning itself to bankroll the robot revolution while simultaneously selling it the shovels. Nice racket if you can pull it off.