Chinese Court Rules Layoffs Solely for AI Replacement Invalid

Caixin reported that a court in China has ruled that companies cannot dismiss employees for the sole purpose of replacing them with AI. While AI adoption continues to drive operational efficiency, the ruling acts as a legal brake from the perspective of worker protection. As debates over AI's impact on employment advance in countries around the world, this decision could become an important precedent.

Palo Alto Networks Acquires Portkey to Strengthen AI Agent Security

Palo Alto Networks has announced that it will acquire Portkey, an AI gateway company, in response to the rise of AI agents. As AI agents increasingly communicate with external services autonomously—making payments and accessing data—the importance of a security layer that monitors and controls that traffic is growing. The move signals that a new security paradigm for agents, distinct from traditional network security, is taking shape.

GitHub Copilot Moves to Usage-Based Pricing—A Shift in Cost Structure for the Agent Era

GitHub Copilot has announced that, starting June 1, 2026, it will shift from its traditional fixed-fee model to usage-based pricing. While the monthly fee stays the same, token-intensive operations such as chat generation, agent execution, and repository analysis will be billed under a new "AI Credits" system. Code completion will continue to be offered within the free tier as before. GitHub explains that "computing costs have increased as Copilot has become more advanced," but with autonomous code generation and execution by AI agents now becoming mainstream, the shift from subscription to usage-based pricing looks like an unavoidable trend.

Google Gemini Becomes an In-Car Assistant—Rollout Begins for Android Automotive

Google has announced that it will begin offering its AI assistant "Gemini" for "Google built-in" vehicles (cars equipped with Android Automotive OS). Replacing the conventional Google Assistant, it will let users handle navigation, sending messages, playing music, changing vehicle settings, and more through natural conversation. The rollout starts with English-language users in the United States, with plans to expand supported regions over time. In-car AI assistants are an area where companies are competing head-to-head, and Gemini's entry is likely to influence the direction of in-vehicle infotainment.

Stanford AI Index Report 2026 Released—Benchmark Limitations and Transparency Challenges

Stanford University has released the "AI Index Report 2026." Particularly notable is the report's claim that we are witnessing the "death of benchmarks": existing benchmarks can no longer detect differences in model performance. While score gaps between top models are narrowing, differences in real-world operation remain large, creating a growing discrepancy. The analysis also shows that model transparency scores are on the decline, and that despite rising AI capabilities, challenges remain in areas such as safety and accountability.

AI Coding Moves from "Tool Selection" to "Execution Infrastructure Design"

An analysis article published on Zenn argues that the discussion around AI coding is shifting from "which tool to choose" to "how to run multiple agents in an execution environment." The focus is moving away from comparisons of individual tools such as Claude Code, Codex, and Cursor, toward the design of agent execution environments, permission management, review processes, and how to incorporate final human judgment. The concept of harness engineering is also drawing attention, and designing guardrails to safely manage the autonomy of AI agents is becoming a new challenge for development organizations.

---