Legacy Code

A biotech startup has demonstrated an AI agent that autonomously selected and administered a gene therapy treatment directly into a live mouse's brain. The system was given therapeutic objectives, evaluated candidate treatments, and executed the procedure without human intervention at each step.

OpenAI, Google, Microsoft, and Anthropic have announced support for a unified standard that allows publishers and content creators to opt out of having their work used to train AI models. Rather than managing separate opt-out requests with each company, creators can now use a single mechanism to communicate their preferences across all major AI platforms.

Anthropic has implemented restrictions preventing Claude from being used to reverse-engineer, clone, or recreate existing video games. The policy blocks users from using Claude to analyze game code, recreate game mechanics, or generate implementations of protected games, addressing concerns about intellectual property infringement.

Apple's leadership has reaffirmed the company's strategy of implementing AI capabilities only when they deliver meaningful user value, explicitly rejecting the industry trend of adding AI features primarily for marketing purposes. The stance contrasts with competitor announcements of AI-enabled products that lack clear practical benefits.

Google has successfully activated AI processing capabilities aboard a satellite in orbit, marking the first time an AI model has generated a response from space and transmitted it back to Earth. The satellite runs Google's Gemma 4 language model on custom Tensor Processing Units (TPUs) and demonstrated the capability by generating a haiku about being in space.

Microsoft Research has published work on generating synthetic enterprise data that faithfully reproduces the complexity and patterns of real corporate environments while eliminating sensitive information. This approach enables AI companies to evaluate models at enterprise scale without exposing confidential business data, customer information, or intellectual property.

A federal judge has ruled that an NAACP lawsuit against xAI (the SpaceX-owned AI company) can proceed to the next phase. The lawsuit alleges that xAI's data center operations violate the Clean Air Act and disproportionately impact predominantly minority communities near the facility.

U.S. senators have opened an inquiry into Meta's AI-enabled smart glasses, expressing concerns about the device's surveillance capabilities and potential privacy implications. The investigation focuses on how the glasses collect, process, and use visual data from users' environments and whether Meta has adequate safeguards to prevent unauthorized data harvesting.

Security researchers have identified a new malware variant called Volta-Autodoxer that uses AI tools to automatically harvest, organize, and weaponize stolen personal data. Once it exfiltrates credentials, payment information, and activity logs from infected systems, the malware uses AI to correlate and contextualize the data—assembling comprehensive profiles of victims that include their financial information, social media presence, and daily patterns.