noticias ia para empresas microsoft copilot ciberseguridad

Default settings are once again making decisions for companies that don't make their own: Microsoft will switch on Anthropic's Claude models in Copilot for every user at European organizations that haven't picked a setting by December 9. Your AI supply chain is also your risk chain: a model provider that works for Notion has confirmed that attackers stole customers' billing data and credentials. And the money now has a price tag: Anthropic is asking $70 to $78 a share and plans to start trading on November 19, with Microsoft and Google among the companies buying shares on the side. This follows Tuesday's radar.

Microsoft will turn on Claude in your company's Copilot on December 9: if you don't choose first, Microsoft chooses for you

Microsoft has told Microsoft 365 admins that starting December 9, Claude models will be enabled for all users at EU and EFTA organizations that haven't made a selection in the AI providers setting. Until now they've been off by default in Europe, and from early December they'll be processed inside Microsoft's EU Data Boundary, with Anthropic as a subprocessor. Microsoft is also retiring the option to allow them only inside Word, Excel and the other Microsoft 365 apps: tenants on that option move to "All users," including in Copilot Studio, where agents get built. For your business: before December 9, go to the admin center (Copilot, Settings, AI providers operating as Microsoft subprocessors, Anthropic) and pick "No users" or a specific group; you need the AI Administrator or Global Administrator role. If you leave it on, add Anthropic to your GDPR subprocessor records. A new model handed to the whole staff without anyone deciding it isn't a company at the INTEGRATED rung: it's a company whose tools get swapped without anyone being told. Source

Billing data and credentials stolen from a model provider that works for Notion: your vendors have vendors too

Fireworks AI, which serves AI models to more than 10,000 customers, has expanded its notice on an attack it detected on October 6: on top of credentials stored on its platform, the attackers took account and billing data and, for some customers, datasets and partially trained models. On October 24, an intruder reached its development machines through the cloud provider hosting them. Notion, which used Fireworks to classify data for an analytics feature and to serve open-weight models, learned of the incident on November 2, cut Fireworks off, deleted its data there and switched that feature off. Fireworks ties the attack to a group that resells stolen access to AI models, what Google and Mandiant call LLMJacking: in their investigations this year they've seen fraudulent charges of more than $100,000 in a single day and up to $1 million in total before access was cut. For your business: ask your AI-powered tools for the list of model providers they use and how many days they take to notify you of an incident; if you use your own API keys, put a spending cap and a daily alert on them, and rotate any that passed through a third party. Source

Anthropic goes public on November 19 at $70 to $78 a share: Microsoft and Google are buying stock in the model they sell you

Anthropic has filed an updated prospectus: it's offering 825 million shares at $70 to $78, which at the top of the range would raise up to $64.35 billion and value it at up to $1.405 trillion, below the $1.5 trillion its own investors were doubting last week. According to people who spoke to CNBC, it will price on the 18th and start trading on the Nasdaq on the 19th. AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC will separately buy $26 billion of stock at the IPO price. The filing puts its annualized revenue at roughly $81 billion at the end of October, $10 billion more than a month earlier. For your business: Microsoft and Google, which sell you Claude inside Copilot and their cloud, now have a bigger stake in you using it; our read is that your platform's "default" model is also a commercial decision, so compare it on cost per finished task with your own tests. Anthropic already warned in its prospectus that prices, limits and models will change: with a company adding $10 billion in a month, make sure your process can switch models without being rebuilt. Source

What to watch tomorrow

Whether more companies that used Fireworks notify their own customers, as Notion did, and whether Anthropic moves its price range during the investor roadshow. Background: here's how we covered its first prospectus and investors' doubts.