OpenAI Builds Secret AI Super-Hacker While Chinese Rivals Challenge U.S. Dominance and Meta Eyes $10B Data Center Deal

OpenAI secretly develops GPT-Red, an AI super-hacker capable of novel prompt injection attacks, while Chinese rivals Moonshot and Alibaba unleash powerful low-cost models challenging U.S. dominance, and Meta negotiates a staggering $10 billion data center deal with Anthropic as the global AI arms race intensifies.

• OpenAI has developed GPT-Red, a powerful LLM super-hacker trained through self-play to identify vulnerabilities in AI models, including a novel 'fake chain of thought' prompt injection attack, though the company refuses to release it publicly over security concerns.
• Chinese AI firms Moonshot and Alibaba are aggressively challenging U.S. AI dominance with massive open-source and open-weight models like Kimi K3 and Qwen3.8, offering competitive capabilities at significantly lower costs and intensifying the global AI race.
• Meta and Anthropic are in negotiations for a data center deal worth up to $10 billion, signaling a major push to strengthen AI infrastructure as computational demands surge across the industry.