Ubuntu/Canonical Infrastructure Down for Over 24 Hours Amid DDoS Attack

Ubuntu's server infrastructure, operated by Canonical, has been under a massive DDoS attack since the morning of May 1 local time and remains unrestored more than 24 hours later. A pro-Iranian government group has claimed responsibility, and the attack is believed to be a sustained, cross-border campaign carried out with the DDoS attack service "Beam".

While updates continue to be served normally from mirror sites, Canonical's web pages and Ubuntu's official download servers remain inaccessible. Of particular concern, the disclosure of a critical Linux vulnerability (root privilege escalation) is proceeding in an incomplete form while official communication from Canonical stalls.

Pentagon Signs AI Contracts with OpenAI, Google, and Others — Anthropic Left Out

It has been reported that the U.S. Department of Defense has signed major contracts with leading AI companies. While AI firms including OpenAI and Google are listed as partners in defense cooperation, Anthropic was reportedly excluded from the deals. As Pentagon AI procurement ramps up in earnest, the move has thrown into sharp relief the differing stances of these companies toward military use.

Anthropic has long positioned itself as safety-focused, and the exclusion appears consistent with the company's policy. As the defense AI market continues to expand, the differences among AI companies' positions are becoming increasingly clear.

NVIDIA Releases "Nemotron 3 Nano Omni," a Compact Multimodal Model with Speech Input Support

NVIDIA has released "Nemotron 3 Nano Omni," the latest version of its multimodal model series. It is the first in the series to natively support speech input, and its standout feature is the ability to jointly process four modalities: text, images, video, and audio.

Compared with the previous-generation Nemotron Nano V2 VL, it achieves consistent accuracy improvements across all modalities, which NVIDIA attributes to architectural refinements along with improved training data and recipes. The model is expected to be deployed on edge devices and in real-time applications.

Medical AI Begins to Outperform Physicians in Diagnostic Accuracy

According to the journal Science, multiple studies now report that AI systems are beginning to surpass physicians in diagnostic accuracy. This marks a significant turning point, showing that clinical AI adoption is shifting from the theoretical stage to practical deployment.

At the same time, Google DeepMind has unveiled its "AI co-clinician" concept, positioning AI as a collaborator for physicians and setting out to build a new model of healthcare. AI's role is increasingly being defined not as a replacement for doctors, but as a complementary tool.

LlamaIndex CEO: "The AI Scaffolding Layer Is Collapsing"

The CEO of LlamaIndex argues that the "scaffolding layer" underpinning AI application development is collapsing. The analysis goes that as LLM capabilities improve, much of the layer that has so far been patched together with RAG pipelines, agent frameworks, and the like is becoming unnecessary.

What survives will likely be tools specialized in domains that LLMs alone fundamentally cannot solve, along with optimization layers that draw out maximum model performance. The AI development tooling ecosystem is facing a major turning point.

---