A study using the SCONE-bench benchmark of 405 blockchain smart contracts found that Claude Opus 4.5, Claude Sonnet 4.5, and GPT-5 collectively developed exploits valued at $4.6M.
The controversy landscape in 2025 was dominated by AI labor displacement and military AI ethics. Mass layoffs at tech companies while AI spending surged created a narrative tension that sustained coverage for months. OpenAI's pivot to defense and Anthropic's resistance to Pentagon pressure became the defining ethical debate of the year.