According to The Information: A memo submitted by Anthropic employees outlining nearly 50 research projects reveals that the company’s focus is on malicious AI agents and deceptive models.