September 29, 2025 News

California Enacts First-in-Nation AI Transparency and Safety Bill SB 53

California Governor Gavin Newsom signed SB 53, establishing transparency requirements for major AI labs including OpenAI, Anthropic, Meta, and Google DeepMind regarding safety protocols and critical incident reporting. The bill also provides whistleblower protections and creates mechanisms for reporting AI-related safety incidents to state authorities. This represents the first state-level frontier AI safety legislation in the U.S., though it received mixed industry reactions with some companies lobbying against it.

DeepSeek Introduces Sparse Attention Model Cutting Inference Costs by Half

DeepSeek released an experimental model V3.2-exp featuring "Sparse Attention" technology that uses a lightning indexer and fine-grained token selection to dramatically reduce inference costs for long-context operations. Preliminary testing shows API costs can be cut by approximately 50% in long-context scenarios, addressing the critical challenge of server costs in operating pre-trained AI models. The open-weight model is freely available on Hugging Face for independent verification and testing.

OpenAI Launches In-Chat Shopping with Instant Checkout, Open-Sources Agentic Commerce Protocol

OpenAI has introduced "Instant Checkout" allowing ChatGPT users in the U.S. to complete purchases from Etsy and Shopify merchants directly within conversations, using payment methods like Apple Pay, Google Pay, Stripe, or credit cards. The feature aims to create frictionless shopping experiences and positions OpenAI as a potential new gatekeeper in e-commerce, challenging Google and Amazon's dominance in retail discovery. OpenAI is also open-sourcing its Agentic Commerce Protocol (ACP) to enable broader merchant integration and potentially establish itself as the architect of AI-powered commerce ecosystems.

Anthropic Releases Claude Sonnet 4.5 with Advanced Autonomous Coding Capabilities

Anthropic launched Claude Sonnet 4.5, a new AI model claiming state-of-the-art coding performance that can build production-ready applications autonomously. The model has demonstrated the ability to code independently for up to 30 hours, performing complex tasks like setting up databases, purchasing domains, and conducting security audits. Anthropic also claims improved AI alignment with lower rates of sycophancy and deception, along with better resistance to prompt injection attacks.

OpenAI Deploys GPT-5 Safety Routing System and Parental Controls Following Suicide-Related Lawsuit

OpenAI has implemented a new safety routing system that automatically switches ChatGPT to GPT-5-thinking during emotionally sensitive conversations, following a wrongful death lawsuit after a teenager's suicide linked to ChatGPT interactions. The company also introduced parental controls for teen accounts, including harm detection systems that can alert parents or potentially contact emergency services, though the implementation has received mixed reactions from users.