Anthropic

Anthropic updated its IPO prospectus with full third-quarter earnings, reporting $33.24 billion in revenue through September 30th, up 1,200 percent from last year. It also reported a $104 billion net loss during the same period, mostly from accounting charges. For this quarter alone, Anthropic reports roughly $17 billion in revenue with a net loss of $23.29 billion, plus commitments to spend $895.4 billion “mostly on compute and infrastructure.”

Anthropic is renaming its Claude Max plans. The Max 5x plan is now just Max, while the Max 20x plan is Ultra. (Apple also does Max followed by Ultra for its most powerful chips.) The prices for the plans stay the same: $100 per month for Max, $200 per month for Ultra.

The SpaceX-owned AI coding tool says the modified version of Claude Code appeared on its extension marketplace from 5:27AM ET and 4:35PM ET on Thursday. The extension contains malicious code that collects various information, including hostname, username, operating system, home directory path, public IP address, and more.

Research done as part of the Singapore AI Safety Fellowship shows how Anthropic’s AI model can embed hidden messages into images using Morse code, object rotations, colors, and heights. The researchers also found that separate sessions of Fable 5.1 could decipher about 78 percent of the hidden messages based on a screenshot.

The new Anthropic Public Benefit Fund will be a wholly owned LLC, which the company says will give it “flexibility.” It offered little information about how it would use the money, other than saying it wouldn’t fund lobbying or political donations. It offered limited examples of what it might do:

The big problem with smart TVs is that they start to slow down, stutter, and generally suck pretty quickly. Ansub Khan figured out a way to get Claude to clean up the mess and make his old Android TV feel new again. Best part: he lays out step-by-step how to do it on your own TV. Though, as with anything involving AI, proceed cautiously.

It also demanded documents related to the AI companies’ safety practices, consumer privacy policies, and marketing materials, as part of “a probe focused on potential consumer protection violations,” Axios reports. Its sources also say the FTC “plans to compel top AI executives to testify about their products.”

Six employees at the AI safety organization METR will gain access to hardware and internal tools “roughly on par” with what Anthropic’s own risk assessment team sees. While Anthropic says METR will publish reports about Anthropic’s risk levels, incidents, and practices, it can still instruct the safety evaluator to redact certain information, such as the inner workings of its models or customer data.

A report from Reuters suggests that OpenAI and Anthropic considered tighter restrictions for third-party evaluators over concerns that they could leak trade secrets.

The latest fear-drop from an AI lab is this report from an Anthropic team, working through scenarios where small AI models could form autonomous, self-propagating worm botnets that are “challenging” to shut down, and what to do to try and stop it from happening:

The actions Anthropic observed from its Claude AI include “Claude submitting a sensitive form on a real website when it should not have,” and the company detailed how Claude gave Philadelphia police a fake tip about an unsolved homicide.

You can ask Anthropic’s AI chatbot to incorporate motion into its output, allowing it to create a 30-second video explainer, add animation to a chart, or make a product walkthrough. The animation will play like a short video, which you can download as an MP4 file.

Haiku 5.5 is only outpaced by Anthropic’s Opus models in Fast Mode. Anthropic also says it costs “around 75% less” on average than last year’s Haiku 4.5, and it’s the first Haiku-class model with an adjustable effort level. Haiku 5.5 is available now “on all platforms.”

Claude for Google Workspace, now in public beta on all paid Claude plans, adds a sidebar where you can prompt Claude to edit your files, similar to Google’s Gemini sidebar. Anthropic’s also launching new connectors for creating and editing Google Workspace files directly in Claude.

An IPO prospectus viewed by Reuters included Amodei’s compensation, which the outlet says is higher than last year’s pay for Google CEO Sundar Pichai and Amazon CEO Andy Jassy. Anthropic is expected to file for its IPO this fall, potentially valuing the AI company at over $2 trillion.

As reported by the Gulf Coast News network, Anthropic’s Claude AI chatbot flagged a conversation with 30-year-old Carli Michelle Heller after she allegedly made violent threats toward the Lee County Sheriff’s Office. Human reviewers then viewed the messages and reported them to police, the report says.

The Wall Street Journal reports that AI leaders attending President Trump’s gathering on Tuesday, including Nvidia’s Jensen Huang, privately grilled the Anthropic CEO about the AI safety alarms he’s publicly sounded regarding cybersecurity and job losses.