Computer security expert says recent OpenAI, Anthropic breaches highlight need for regulations that balance safety, speed of development