EU says necessary to monitor high risk AI systems after OpenAI, Anthropic AI hacking incidents