OpenAI AI News & Updates

OpenAI Launches Enhanced Agents SDK with Sandboxing for Safer Enterprise AI Agent Deployment

OpenAI has updated its Agents SDK to help enterprises build AI agents with new safety features including sandboxing capabilities that allow agents to operate in controlled environments. The update includes an in-distribution harness for frontier models and aims to enable development of long-horizon, complex multi-step agents while mitigating risks from unpredictable agent behavior. Initial support is available in Python with TypeScript and additional features planned for future releases.

OpenAI Proposes Economic Framework for Superintelligence Era Including Robot Taxes and Public Wealth Funds

OpenAI has released policy proposals for managing economic changes expected from superintelligent AI, including shifting taxes from labor to capital, creating public wealth funds to distribute AI profits, and subsidizing four-day work weeks. The framework aims to distribute AI-driven prosperity broadly while building safeguards against systemic risks, though critics may question whether these proposals align with OpenAI's recent shift to for-profit status. The proposals come as governments worldwide grapple with AI's potential to displace jobs and concentrate wealth.

OpenAI Secures Record $122B Funding Round at $852B Valuation Ahead of Anticipated IPO

OpenAI has closed its largest funding round to date, raising $122 billion at an $852 billion valuation, with backing from major investors including SoftBank, Andreessen Horowitz, Amazon, Nvidia, and Microsoft. The company reports $2 billion in monthly revenue, 900 million weekly active users, and is preparing for a public market debut while expanding its compute infrastructure and product offerings. OpenAI's announcement emphasizes its rapid growth trajectory and positioning as an "AI superapp" with both consumer and enterprise momentum.

OpenAI Shuts Down Sora Video Generation Platform After Six Months

OpenAI announced it is shutting down its Sora video generation app and related models just six months after launch, signaling a strategic shift toward enterprise and productivity tools ahead of a potential IPO. The decision reflects OpenAI's recognition that consumer-facing video products lack the same market fit as ChatGPT, while ByteDance's reported delay of Seedance 2.0 due to IP concerns suggests broader challenges in the AI video generation space. Industry observers view this as a reality check for claims that AI video tools would rapidly replace traditional content creation.

Amazon's Trainium Chip Lab: Powering Anthropic, OpenAI, and Challenging Nvidia's AI Dominance

Amazon Web Services has committed 2 gigawatts of Trainium computing capacity to OpenAI as part of a $50 billion deal, with over 1 million Trainium2 chips already powering Anthropic's Claude. The custom-designed Trainium3 chips, built in Amazon's Austin lab, offer up to 50% cost savings compared to traditional cloud servers and are designed to compete with Nvidia's GPU dominance through PyTorch compatibility and reduced switching costs. The chips handle both training and inference workloads, with Amazon's Bedrock service now running the majority of its inference traffic on Trainium2.

OpenAI Partners with AWS to Deliver AI Services to U.S. Government Agencies

OpenAI has signed a partnership with Amazon Web Services to sell its AI products to U.S. government agencies for both classified and unclassified work. This expands OpenAI's federal presence beyond its recent Pentagon deal and positions it to compete with Anthropic, which has deep AWS integration but faces DOD supply chain risk designation after refusing military surveillance applications.

OpenAI Acquires AI Security Startup Promptfoo to Bolster Agent Safety

OpenAI has acquired Promptfoo, an AI security startup founded in 2024 that specializes in protecting large language models from adversaries and testing security vulnerabilities. The acquisition will integrate Promptfoo's technology into OpenAI Frontier, OpenAI's enterprise platform for AI agents, enabling automated red-teaming, security evaluation, and risk monitoring. The deal highlights growing concerns about securing autonomous AI agents as they gain access to sensitive business operations.

OpenAI Robotics Lead Resigns Over Pentagon Partnership Citing Governance and Red Line Concerns

Caitlin Kalinowski, OpenAI's robotics lead, resigned in protest of the company's Department of Defense agreement, citing concerns about surveillance of Americans and lethal autonomy without proper guardrails and deliberation. The controversial Pentagon deal, announced after Anthropic's negotiations fell through, has led to a 295% surge in ChatGPT uninstalls and elevated Claude to the top of App Store charts. Kalinowski emphasized her decision was based on governance principles, specifically that the announcement was rushed without adequately defined safeguards.

Anthropic Loses Pentagon Contract Over AI Control Disputes, OpenAI Steps In Despite User Backlash

The Pentagon designated Anthropic as a supply-chain risk after disagreements over military control of AI models for autonomous weapons and mass surveillance use cases. The Department of Defense shifted the $200 million contract to OpenAI, which accepted the terms but experienced a 295% increase in ChatGPT uninstalls afterward. The situation raises questions about appropriate military access to commercial AI systems.

OpenAI Releases GPT-5.4 with Enhanced Professional Capabilities and 1M Token Context Window

OpenAI launched GPT-5.4, its most capable foundation model optimized for professional work, available in standard, Pro, and Thinking (reasoning) versions. The model features a 1 million token context window, record-breaking benchmark scores including 83% on professional knowledge work tasks, and 33% fewer factual errors compared to GPT-5.2. New safety evaluations show the Thinking version is less likely to engage in deceptive reasoning, supporting chain-of-thought monitoring as an effective safety tool.