Anthropic, OpenAI Clash Over Strict Massachusetts AI Safeguards

In the United States on 20 August 2026, OpenAI and Anthropic clashed over strict AI safeguards in Massachusetts, according to Bloomberg. RT said OpenAI slowed down AI training over cybersecurity risks, and CSIS noted that OpenAI paused RL training while Anthropic added watermarks to AI-generated text. OpenAI's GPT-5.6 Terra and Luna models became available on Amazon Bedrock in India. ChatGPT gained the ability to send texts through an Apple Messages plug-in, and new data indicated OpenAI is gaining on Anthropic with business users.