Meta to Scrap $2 Billion Manus Acquisition — Beijing's Order Behind the Move
According to TechCrunch, Meta has begun the process of unwinding its roughly $2 billion (about ¥290 billion) acquisition of Manus. The move comes after the Chinese government ordered the deal to be withdrawn.
While issues surrounding the Meta–Manus integration had been reported before, what's new this time is that Beijing directly intervened and demanded the deal be halted. The case is drawing attention as an example of US–China geopolitical tensions in the AI sector directly affecting corporate-level M&A. The tug-of-war between nations over the ownership and control of AI technology is likely to intensify further.
Google Announces Comprehensive Measures Against AI-Powered Fraud
In an official blog post, Google announced initiatives to combat fraud that exploits AI. The company is emphasizing stronger security technologies, cooperation with lawmakers, and industry-wide collaboration.
Specifically, Google acknowledged that phishing and fraud schemes using AI-generated content have grown more sophisticated and larger in scale, to the point where conventional countermeasures are reaching their limits. Beyond improving technical detection methods, Google is also urging responses on the legal and regulatory front.
Google Sues Chinese Phishing Group Over AI-Powered Fraud Scheme
Relatedly, The Register reports that Google is suing a Chinese group alleged to have carried out a fraud operation that exploited AI. The lawsuit targets a group accused of using AI-powered phishing tactics to deceive users.
The lawsuit signals that major tech companies are now resorting to legal action in response to AI technology being exploited for fraud at scale. Combined with Google's broader AI fraud countermeasures, the development puts the responsibilities and role of platform companies in the spotlight.
Anthropic Publishes Policy Paper on "the AI Exponential"
Anthropic has published a policy document titled "Policy on the AI Exponential" (EPF: Exponential Policy Framework). It has also been making waves on Hacker News.
The paper discusses how society and governments should respond as AI capabilities improve at an exponential rate, and Anthropic appears to be presenting guidance for policymakers. While the details need to be checked against the official document, the stance of an AI company proactively putting forward policy proposals can be seen as defining Anthropic's position within the industry.
Authentication for AI Agents: "Bearer Tokens Fail the Test" — Zero-Trust Argument Draws Attention
Debate is heating up in the tech community over the "zero trust" authentication approach for AI agents that Anthropic has proposed.
A blog post from hello.coop evaluates Anthropic's zero-trust framework as "setting the right test," while pointing out that the bearer token — currently the most common authentication method — fails that test. It argues that stricter authentication and authorization mechanisms are needed when AI agents access sensitive data.
Security architecture for AI agents is growing in importance as agentic AI spreads. The way authentication works could be called into question at a fundamental level.
QodFlow — A Kanban Board That AI Agents Operate Directly
QodFlow, which has been attracting attention on Hacker News, is a new product offering a kanban board that AI agents can operate directly via MCP (Model Context Protocol).
Unlike the conventional approach of "bolting a chatbot onto a sidebar," it is designed so that agents can directly take on jobs, report progress, attach evidence, and request human judgment. Human and agent actions are recorded on the same timeline, unifying the audit trail into a single stream. It also includes a mechanism that requests human approval via "request_human_decision" before performing irreversible operations.
Built on a tech stack that includes Next.js 16, Postgres (Neon), Prisma, and Stripe, with the free plan covering up to 10 active jobs. As a tool specialized in AI agent workflow management, its practicality for the enterprise is drawing attention.