Anthropic's New Models "Fable 5" and "Mythos 5" Pulled from Release by the US Government
Over the past weekend, the US government asked Anthropic to withdraw the release of its latest models, Fable 5 and Mythos 5, citing national security concerns. Reports indicate the move came after Amazon's research team discovered techniques for bypassing Fable 5's guardrails.
In response to the decision, cybersecurity researchers published an open letter calling it "dangerous." Anthropic itself has pointed out that similar jailbreak techniques exist in other models as well. Meanwhile, despite the ban, reports suggest that Anthropic's brand recognition and market interest have actually grown, underscoring the complex relationship between regulation and brand impact.
This case is likely to put AI model safety evaluation processes and the design of national-level AI regulation at the center of upcoming debates.
EU to Build an Open-Source AI Model Covering All 24 Official Languages — EUROPA Consortium Selected
The European Commission has selected "EUROPA," a European consortium led by Italy's Domyn, as the winner of the Frontier AI Grand Challenge. The project aims to build an open-source frontier AI model that supports all 24 official EU languages.
The requirements call for a model of at least 400 billion parameters — a scale on par with the world's most advanced AI systems. The model will be released openly and developed on infrastructure within Europe, positioning it as a symbol of European "technological sovereignty."
European Commission Vice President Henna Virkkunen stated that "Europe can lead advanced AI on its own terms," emphasizing an AI development approach centered on openness, trust, and strategic autonomy. The challenge, which opened for applications in February 2026, represents a major step toward Europe developing its own AI capabilities.
Norway Effectively Bans AI Use in Elementary Schools
Norway has introduced regulations that substantially ban AI use in elementary schools. While debates over AI in education continue in many countries, Norway's approach stands out as particularly strict. Concerns about impacts on children's learning development and privacy appear to be behind the move, though the specific details and exception provisions could change significantly depending on how the rules are applied going forward.
59% of the Feed for New TikTok Users Is AI-Generated Content
According to research by Kapwing, 59% of the videos in the "For You Page" feed served to new TikTok users consist of AI-generated content (AI slop). This suggests that the platform's recommendation algorithm may be prioritizing AI-generated content for distribution.
The figure illustrates how the democratization of content creation is affecting platform ecosystems in unintended ways. As it becomes increasingly difficult for users to distinguish human-made content from AI-generated content, platforms are facing growing scrutiny over how they respond.
Open Models' Cost Competitiveness Improving Rapidly
One point drawing attention in community discussions is that the cost-effectiveness of open-weight models is rapidly catching up with closed-source models. Models such as DeepSeek, Qwen, GLM, Kimi, and MiniMax are filling the "good enough and cheap" niche across many real-world use cases.
What's particularly interesting is the observation that the cost gap between frontier models and strong open models has grown far wider than the performance gap between them. API-based closed-source models still offer advantages such as no infrastructure requirements and high reliability, but for use cases that prioritize privacy, customizability, and cost predictability, open models are becoming a compelling option.
Reliance to Integrate AI Across Its Telecom Service with 500 Million Users
India's Reliance Industries is set to embed AI across its telecom service, which is used by more than 500 million people. Chairman Mukesh Ambani has indicated plans to integrate AI into every touchpoint, from calls and apps to home services.
This is not merely an experiment — it is a noteworthy real-world example of layering AI on top of existing large-scale infrastructure. It suggests that AI adoption in emerging markets may proceed at a different scale and in different forms than in developed countries.
---