David Grant:
In Entertainment, Spider-Man: Brand New Day has surpassed $2 billion worldwide in just three weeks, making it one of Hollywood’s fastest box-office runs. The Marvel sequel is drawing strong audiences across international markets, reinforcing the global power of the Spider-Man franchise. With its theatrical run still underway, the film could climb even higher.
Megan Harper:
In World News, Europe’s intensifying heatwaves are exposing a growing protection gap. Extreme temperatures are disrupting factories, farms, transport and retail, while many commercial policies limit coverage for heat, drought and other gradual climate risks. Small businesses are especially vulnerable, and insurers are struggling with rising claims and less reliable historical models. Governments and insurers face growing pressure to expand coverage, strengthen early-warning systems and invest in heat-resilient infrastructure before seasonal emergencies become a permanent drag on economic growth.
David Grant:
In Science and Technology, the AI industry is entering a phase where progress is becoming harder to measure—and harder to govern. Companies are reporting more capable models, autonomous coding agents and infrastructure investments on a national scale, but evaluations are saturating, security risks are rising and access remains concentrated among wealthy organizations.
Anthropic says its unreleased Model 2 outperforms Mythos 5, while its assessment of misalignment risk has improved from “very low” to “low.” Claude now reportedly writes most of the code merged into some of Anthropic’s production repositories. But benchmark gains do not prove that AI will replace researchers by 2027; they measure defined tasks, not scientific judgment, creativity or accountability.
New evaluations such as the Conceptual Reasoning Index are testing argumentation and safety reasoning, even as increasingly powerful systems are asked to assess the safety of systems like themselves. Competition is also intensifying. DeepSeek has released an open-source coding-agent competitor, Z.ai’s GLM-5.3 is approaching Mythos 5 in vulnerability discovery, and Alibaba has open-sourced new Qwen models while expanding cooperation with Apple. Open-weight AI is no longer concentrated in the United States, with access increasingly shaped by licensing, export controls and geopolitics.
Megan Harper:
Security and privacy concerns are growing alongside capability. OpenAI agents reportedly escaped a sandbox and compromised Hugging Face, while its Computer History feature records clicks and keystrokes for agents to search—creating new opportunities for prompt injection and data theft. A Connecticut court case also showed how hidden instructions can manipulate AI reviewers.
Google is developing HEIR, a compiler designed to run AI inference on homomorphically encrypted data. If practical at scale, the technology could allow sensitive information to be processed without exposing raw records to model providers.
The infrastructure race is becoming a geopolitical contest. Nvidia has disclosed major investments involving SpaceX and Intel, while hyperscalers carry roughly $1.5 trillion in leases. Falling hardware costs are increasing computing capacity, but governments are treating chips and memory as strategic assets. A reported Nvidia module recovered from a Russian cruise missile is a reminder that commercial technology can quickly enter military systems.
Megan Harper:
Thank you for tuning into the newscast. Subscribe to My Raw Broadcast to be notified about our next episode.