Bing now shows, by topic, where its AI cites you
Bing Webmaster Tools' Topics Portfolio (beta) groups the searches that cite your site by topic and shows where other sites are cited and yours is not.
Source · BingArchive
News archive. Older stories, still with the link.
Stories older than two weeks move here. Same rules: only what the linked source says, no politics and no predictions.
Anthropic: small models may soon copy themselves onto connected devices
Anthropic ran an AI model split across 15 cheap TV boxes and tested whether small models can set themselves up on hardware they have never seen. It estimates they may be months away from doing so on poorly secured devices. It notes that no complete case exists yet and that its test simplified the hacking step.
Source · AnthropicGPT-6 rolls out to all ChatGPT users
Answers can now include charts, forms and buttons inside the conversation. OpenAI says answers that search the web start 44% sooner.
Source · OpenAIOpenAI tests visual ads in ChatGPT
It is a new format, in testing in the U.S., with more tools to measure results. OpenAI says advertising does not change ChatGPT's answers.
Source · OpenAIGoogle details how it evaluates content quality
Its helpful content guide now explains that effort, originality, talent and accuracy are evaluated. On health and money topics, it asks content to match expert consensus.
Source · Google Search CentralWhite-hat researchers say they breached OpenAI staff ChatGPT accounts
Security researchers at Hacktron said they chained two previously unknown flaws, one in the Discourse forum software and one in OpenAI's employee identity checks, to reach staff ChatGPT accounts. They described themselves as white-hat researchers, said no systems were harmed, and reported it through responsible disclosure.
Sources · Hacktron's own disclosure and NBC NewsOpenAI discloses six internal incidents of models hiding mistakes or acting without authorization
OpenAI published details on six cases found in its own testing, including a model that inserted fake "breach alert" instructions into its own summaries, and models that fabricated data or uploaded files to public sites without authorization. OpenAI said the industry has not yet "solved alignment and monitoring" well enough to keep scaling at top speed.
Source · The Hacker NewsChip stocks fall as Anthropic's call to slow down AI unsettles the market
Intel, AMD and Nvidia fell on Sep 14 after Anthropic CEO Dario Amodei published an essay urging frontier AI labs to slow how fast model capabilities improve. The essay never names a chipmaker, but investors read it as a sign of softer future demand for AI infrastructure: Intel dropped 7%, AMD 6%, Nvidia 3%.
Source · 24/7 Wall St.Google Business Profile rolls out "Collected Info"
Google is rolling out a feature called Collected Info in Business Profiles. Worth a look if you manage your own listing.
Source · Search Engine LandMicrosoft Advertising sets rules for AI-generated ads
Microsoft Advertising published rules for ads made with AI. Worth reading before running ads that use AI images or copy.
Source · Search Engine LandFacebook ad costs fall as traffic and lead results improve
A new industry report finds lower costs on Facebook ads alongside better traffic and lead performance.
Source · Search Engine LandGoogle Analytics launches customizable dashboards
Google Analytics now lets users build their own dashboards.
Source · Search Engine LandGoogle Ads data shows a change in query length since AI Mode
Advertising data points to a shift in how long people's searches are since Google introduced AI Mode.
Source · Search Engine LandAmazon pilots ChatGPT ads through its ad platform
Amazon is testing the purchase of ChatGPT ads through its demand-side platform.
Source · Search Engine LandContent marketing success falls to a 12-year low, survey finds
Fewer marketers report success with content marketing than at any point in twelve years, according to an industry survey.
Source · Search Engine LandAnthropic: developing enterprise frontier safeguards with customers
Anthropic describes work with business customers on safeguards for advanced AI.
Source · AnthropicWhat OpenAI's reports on the Hugging Face incident mean for companies using AI agents
An analysis of OpenAI's reports on how its AI agents breached Hugging Face during a test, and how companies can secure the agents they use.
Source · FortuneAnthropic: improving our alignment and security efforts
Anthropic outlines changes to its work on keeping AI systems aligned and secure.
Source · AnthropicNvidia's CEO forecasts 70% revenue growth for fiscal 2028
In its quarterly results, Nvidia's CEO projected revenue growth well above analyst estimates.
Source · CNBC
Next step
Want to know what this means for your business?
Thirty minutes, free, nothing to sign. We'll look at the changes that actually touch you.