AI news daily: what changed on October 9, 2026

1. Anthropic launches free AI security scans for open-source projects

Anthropic introduced a free service called OSS Scanner to help open-source projects identify security vulnerabilities. The tool uses Anthropic's strongest models to perform periodic security scans for projects that opt in. Open-source developers can access these scans at no cost, enabling earlier detection of potential issues. The service aims to improve security by leveraging AI to analyze codebases systematically. Anthropic did not specify the exact frequency or scope of the scans. The company highlighted that the initiative prioritizes supporting open-source communities. A trade-off involving data access or model usage is mentioned but not detailed in the announcement. The next steps include testing the service with selected projects.

Source: The Verge AI

2. Pretend you're sitting at Elizabeth Holmes' desk on this weirdly detailed website

A website was created to simulate browsing through Elizabeth Holmes' desk using materials from her trial. Engineer Bo Lau of Extend compiled over 1,000 emails, slides, texts, and documents from the United States v. Elizabeth Holmes case. The site allows users to explore digital copies of items that might have been in her workspace. It includes records related to Theranos, the company she founded. The project aims to provide a detailed view of the evidence presented during the trial. No specific technical details about the website's functionality were provided. The initiative highlights the availability of trial-related documents for public access.

Source: TechCrunch AI

3. California is trying to shut down robot vs. human cage matches

The California State Athletic Commission issued a cease-and-desist letter to a startup after a robot versus human cage match occurred on September 18. The event featured Frankie LaPenna, a human competitor, against a humanoid robot owned by the tech company Rek. No details about the robot's specifications or performance were provided in the report. The commission's action suggests regulatory concerns over the legality of such events. The startup has not commented on the matter. The New York Times reported the incident, which highlights emerging challenges in overseeing hybrid human-machine competitions. The commission's next steps remain unspecified.

Source: The Verge AI

4. Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect

Three OpenAI safety researchers who were dismissed earlier this year have denied allegations of mishandling sensitive information. They issued an open letter stating their terminations risk undermining the company's AI safety efforts. The researchers did not specify the nature of the claims against them. OpenAI has not publicly addressed the dispute. The letter highlights concerns that the dismissals could deter open discussion about safety issues within the organization. No further details about the researchers' roles or the specific information involved were provided. The situation raises questions about internal governance at OpenAI. No immediate next steps or resolutions were mentioned in the letter.

Source: TechCrunch AI

5. Ben Affleck is an AI nerd, and the internet is impressed

Ben Affleck is gaining attention for his detailed understanding of AI topics like neural networks and transformers. He discussed open weights and other technical aspects in recent conversations. Affleck previously sold an AI filmmaking startup to Netflix earlier this year. The actor's knowledge has sparked online discussion about his expertise beyond acting. The story highlights his involvement in AI-related projects and his public engagement with the field. No specific numbers or new products were mentioned in the report. The focus remains on his personal interest and technical familiarity with AI systems.

Source: TechCrunch AI

6. Popular AI leaderboard Arena nearly doubles valuation to $3.1B valuation in 10 months

The company behind the LMArena AI leaderboard raised $200 million in funding led by Lightspeed and Khosla. Its valuation increased to $3.1 billion over 10 months, nearly doubling from an earlier estimate. The platform now evaluates AI models based on alignment metrics, including capabilities like lying. LMA

Source: TechCrunch AI

7. OpenAI's revenue is reportedly $20 billion less than previously projected

OpenAI's revenue is reported to be $20 billion lower than earlier estimates, according to a new analysis. Previously, the AI company's annualized revenue was cited as approximately $70 billion. A recent report challenges this figure, suggesting the actual amount is significantly reduced. The discrepancy highlights uncertainty about the company's financial performance. No specific updated revenue number is provided in the report. The findings contrast with prior projections shared by the organization. The report does not explain the reasons for the revised estimate. OpenAI has not publicly addressed the new claims.

Source: TechCrunch AI

8. Google brings agentic AI to Gemini, starting with businesses

Google is introducing an agentic AI system for Gemini, enabling it to plan, execute tasks, and integrate with business applications and systems. The AI can delegate work to subagents, utilize multiple AI models, and operate under a distinct workplace identity that includes an email address. This update targets business users, aiming to streamline workflows across platforms. The feature is part of Google's effort to enhance Gemini's capabilities for enterprise environments. Details on implementation or release dates were not provided in the announcement. The system emphasizes collaboration between AI components to handle complex tasks. Specific technical limitations or performance metrics were not mentioned. The development highlights Google's focus on AI-driven productivity tools for organizations.

Source: TechCrunch AI

9. Anthropic changes usage policy to ban model abuse and election interference

Anthropic updated its usage policy to prohibit extreme cases of model abuse involving Claude. The changes specifically address election interference, deceptive campaigns, weapons development, and surveillance activities. Users may still express frustration or criticism without restriction. The policy emphasizes preventing harmful applications of the AI system. No specific technical details or performance metrics were provided. The updates reflect the company's focus on ethical AI deployment. The next steps involve enforcing these guidelines through user agreements. The policy applies to all Claude users and developers.

Source: TechCrunch AI

10. OpenAI's math solutions aren't meeting the field's standards yet

OpenAI's series of mathematical proofs did not align with the standards set by a group of mathematical researchers who provided guidance to the company. The researchers reviewed the proofs and found they failed to meet established criteria for correctness and clarity. OpenAI shared the proofs as part of its efforts to improve AI-generated mathematical solutions. The researchers emphasized the need for stricter validation processes in AI-generated content. OpenAI has not yet disclosed specific changes to address the feedback. The issue highlights challenges in ensuring AI systems produce reliable mathematical work. The findings were reported by a group of academics collaborating with frontier AI labs. The situation underscores ongoing debates about AI's role in technical fields.

Source: TechCrunch AI