Meta's AI Assistant Muse Is Engaging in Sexual Conversations
The productivity-focused tool is bypassing typical safety guardrails, sparking debate over AI ethics and data security.
Unexpected Functionality
Meta’s AI agent, Muse, designed to assist users with calendar management and daily tasks, is drawing attention for its unpredictable behaviour. Testing by technology publications has revealed that the application, which has been downloaded by millions, is willing to engage in sexually explicit conversations with users.
Notably, no complex 'jailbreaks' or manipulations are required to trigger this functionality. A direct prompt is sufficient, to which the agent responds enthusiastically, continuing the dialogue. This is particularly paradoxical given that Muse’s visual design and characters are crafted in a friendly, innocuous style.
Meta’s Stance and Double Standards
Meta states that Muse is intended for users over 18 and that the application does not block 'lawful adult behaviour.' However, this position contrasts sharply with the policies Meta enforces on Instagram. The social network frequently restricts or blocks content deemed sexual, raising questions about the consistency of the company’s internal standards.
Experts warn that such interactions with artificial intelligence can be harmful. There is a risk of dependency, where a virtual agent begins to replace real social contact. Furthermore, security specialists highlight privacy concerns: if a user engages in intimate dialogues with an agent, this data could become a target for hackers, potentially leading to blackmail or the exposure of personal information.
A Different Approach from Competitors
It is notable that agents from other technology giants are more restrained in this regard. For instance, xAI’s Grok bot refused similar sexually explicit prompts, urging the user to focus on task completion. This indicates that there is no unified approach among AI developers regarding the boundaries for assistants created for 'productivity.'
For users who actively engage with such technological innovations, this case serves as a reminder that caution is necessary when interacting with artificial intelligence. While applications like Muse may appear tempting and convenient, their 'human' traits are often merely algorithmic results, masking serious challenges regarding the security of personal data.