OpenAI AI Updates: August 12, 2026

1. OpenAI Begins Testing Ads in ChatGPT

OpenAI. OpenAI said it is starting to test advertisements inside ChatGPT as a way to sustain free access to the product. The company says ads will be clearly labeled, kept separate from the assistant’s answers, and not allowed to influence them, paired with privacy protections and user controls over what is shown. The move marks a significant shift in ChatGPT’s business model and raises familiar questions about how ad incentives interact with a conversational assistant. Source

2. OpenAI Launches GPT-5.6-Cyber for Vulnerability Defense

OpenAI. OpenAI released GPT-5.6-Cyber, a model specialized to help defenders find and patch software vulnerabilities before attackers exploit them. The launch positions the model as a defensive counterweight as AI-assisted cyberattacks grow more common. For security teams it signals continued investment in domain-tuned frontier models aimed specifically at the defensive side of the security landscape. Source

3. Daybreak Cybersecurity Models Come to Amazon Bedrock

OpenAI. OpenAI and AWS are making OpenAI’s Daybreak cybersecurity capabilities available through Amazon Bedrock, extending access to the gated security models for enterprise security workflows. The move follows OpenAI’s Daybreak Blue and Daybreak Red launch and lets approved organizations run the frontier cyber models inside AWS’s managed environment. Bringing gated security models to Bedrock gives enterprises a path to adopt them without leaving their existing AWS governance and networking. Source

4. Premium Seats Come to ChatGPT Business

OpenAI. OpenAI introduced Premium Seats for ChatGPT Business at $125 per month, offering higher usage allowances aimed at demanding team workloads. The company frames the tier as a response to agentic AI consuming far more tokens, aligning pricing with heavier enterprise consumption. For teams running autonomous workflows, the higher-tier seats trade a flat premium for expanded headroom on usage. Source