ChatGPT Adds Personal Finance Features — AI Financial Advice with Bank Account Linking

OpenAI has added a new personal finance experience to ChatGPT. Available as a preview for Pro users in the US, it securely links accounts at financial institutions so the AI can deliver personalized insights and advice based on the user's income, spending, and asset situation.

Personal financial advice has long been the domain of human professionals and robo-advisors, but conversational AI like ChatGPT stands apart from existing services by offering "individualized guidance that understands context." That said, open questions remain — from privacy concerns around entrusting financial data to an AI, to how the accuracy of its advice can be guaranteed.

For now this is a US Pro-only preview, with no word on general availability. Since the finance × AI space is subject to strict regulation, a phased rollout seems likely.

Anthropic's Valuation Reaches $900 Billion — Could Overtake OpenAI

Reports indicate Anthropic's valuation is on track to hit $900 billion (roughly 13 trillion yen). If the funding round closes at that figure, it would exceed OpenAI's current valuation.

Anthropic has steadily expanded its enterprise AI market presence with the Claude series, and recently announced an expanded PwC partnership to deploy Claude at a scale of hundreds of thousands of users. Its safety-first approach (Constitutional AI) is widely seen as a good fit for corporate governance requirements.

That said, a valuation spike this sharp is also a sign of overheating across the AI industry as a whole. With limited financial disclosure from these companies, how far profitability actually underpins the numbers remains something to keep a close eye on.

Google Adds "AI Manipulation" to Its Spam Rules

Google has updated its search spam policies, explicitly adding the use of AI to manipulate search results to the list of prohibited practices.

Google has tightened quality standards for AI-generated content incrementally through updates such as the Helpful Content update, but this change is notable for containing language that more directly prohibits "ranking manipulation using AI."

Mass production of AI-generated articles has become the norm in the SEO industry, eroding trust in the search ecosystem. Google's response does not appear to be a blanket ban on AI-generated content; rather, it seems intended to target low-quality AI content that offers no value to users.

Cursor Announces Background Agents

Cursor, the AI editor for developers, has announced a "background agents" feature that autonomously fixes code and executes tasks.

AI coding assistants have traditionally been used interactively — the developer enters a prompt and receives results each time. Background agents take an asynchronous approach: given a task, they autonomously explore the repository, make the necessary changes, and notify you when done.

It's a symbolic release marking the evolution of AI agents from "chat-based" to "autonomous." That said, autonomous code changes make review processes more important than ever.

A Fully Offline Autonomous Conversational Robot Running on Gemma 4

A project for a fully offline autonomous conversational robot running on a Jetson Orin NX SUPER 16GB is drawing attention in Reddit's LocalLLaMA community.

It runs Gemma 4 E4B with Q4_K_M quantization via llama.cpp, achieving a cached TTFT (Time to First Token) of about 200ms and sustained generation at 14–15 tok/s. Speech recognition is handled by SenseVoiceSmall, speech synthesis by Piper, and mouth movements are synced at 43Hz. The face is rendered on the display with PixiJS. More than 30 sensor readings are also woven into the prompt in natural language.

With a fully offline design that has no network interface at all, it's a fascinating example of how AI robots can be used in environments where privacy and security are paramount.

UK Calls on Companies to Address Frontier AI Model Risks

The UK government has issued a statement calling on companies to take measures to mitigate the risks of frontier AI models.

It stops short of concrete regulation, framed as a "should" request — but it makes the government's stance on ensuring the safety of frontier AI (cutting-edge, high-performance AI models) explicit. The UK, also known as the host of the AI Safety Summit, is continuing its approach of encouraging industry self-regulation.