AI issue is a news media that delivers AI news in an easy-to-read way, every day.
Google has revamped its approach to tracking and naming hacker groups. TechCrunch interviewed global experts in hacker tracking to uncover why security companies assign codenames to hacker groups and the practical significance of doing so.
Google·Aug 8, 2026 15:25 UTCMartin Spear, an engineer at OpenAI, has disclosed efforts to automatically manage system performance using AI agents. As agent-based workflows become more prevalent, causing a surge in code changes, OpenAI has implemented a system where always-on AI automatically handles profiling and performance degradation detection, maintaining ChatGPT's speed and stability on a global scale.
OpenAI·Aug 8, 2026 13:50 UTC
SpaceX and Tesla announced plans to jointly build a semiconductor chip manufacturing facility in Texas with a total investment of $16.8 billion. Called 'TeraFab,' the facility is designed to supply chips for AI processing, robotics, and space-based data centers, marking a significant move by companies under Elon Musk's leadership to strengthen in-house semiconductor procurement.
Spacex·Aug 8, 2026 13:36 UTC
OpenAI has revealed that it intentionally slowed the development pace of its new AI model "Astra" after determining that the model had reached the company's internal "critical cybersecurity threshold." The model has been evaluated as capable of autonomously identifying and executing cyber attacks against real-world systems that were traditionally considered to be securely protected, and OpenAI has judged this capability level to be a safety concern.
OpenAI·Aug 8, 2026 13:36 UTC
Anthropic's coding tool Claude Code has gained a new feature enabling multiple sessions to communicate and share information with each other. Running on macOS and Linux, it now allows message transmission and status checks between concurrently active sessions.
Anthropic·Aug 8, 2026 13:33 UTC
Researchers from Coral AI Labs and multiple universities announced 'AgentRadio,' a mechanism enabling multiple AI agents to collaborate in real-time. In a benchmark targeting actual software repositories, a team of 4 agents using AgentRadio demonstrated accuracy rates exceeding those of high-performance standalone models (Claude Opus 4.8). This research outcome illustrates that agent collaboration design has a significant impact on performance, beyond simply enhancing the model itself.
Framework, a PC manufacturer, has notified all customers that hackers illegally obtained personal data including names, email addresses, phone numbers, and addresses of its entire customer base. The leaked information poses a risk of secondary damage such as phishing and impersonation attacks.
Frame·Aug 8, 2026 13:32 UTC
Pangram, a company specializing in AI-generated text detection tools, has announced its latest model, "Pangram 4." According to the company, the model can detect AI-generated text with 99.66% accuracy, with false positives occurring only once in every 24,000 cases. The company also revealed plans to increase API usage fees by 2 to 10 times the previous rate.
Pangram·Jul 29, 2026 17:21 UTC
The U.S. Federal Communications Commission (FCC) has announced measures to regulate the import of Chinese-made humanoid robots and power infrastructure equipment essential for artificial intelligence development. As part of the Trump administration's policy to reduce dependence on Chinese technology, the regulation targets two fundamental fields for the AI industry: robotics and power infrastructure.
DoorDash, a major US food delivery company, has obtained approval from the Federal Aviation Administration (FAA) to operate commercial drone delivery services. This positions the company to formally launch its own drone delivery operations within the United States.
Doordash·Jul 29, 2026 13:20 UTC
Snowflake announced 'Cortex AI Gateway,' a new capability for centrally managing AI agent access to enterprise data and tools. The feature can control access from external AI agents—including those from competitors like Anthropic's Claude Code and Cursor—and was accompanied by the announcement of partnerships with five security companies: 1Password, Aembit, Linx Security, SailPoint, and Saviynt. The initiative aims to update traditional security models designed around human users, as organizations enter an era where AI agents operate on behalf of humans.
Snowflake·Jul 29, 2026 09:20 UTC