Meta’s Muse AI Assistant: Privacy and Cybersecurity Concerns

Meta’s Muse AI assistant has gained attention due to its dangerous implications for privacy and cybersecurity. With over 3 million downloads, it handles tasks like canceling unwanted subscriptions by accessing personal bank accounts, credit cards, emails, and texts. This has raised alarms about financial security. Liz Peek emphasizes the need for robust cybersecurity because of these valid concerns.

AI assistants are being developed to manage a variety of tasks across apps, websites, and accounts. Meta’s Muse and Instinct’s AI agents push this concept further, with Instinct enabling phone calls to businesses for early-access users. Meta’s Muse is testing a similar outbound-calling feature. While convenient, this raises questions about allowing software to make decisions on your behalf.

“Are users comfortable letting software speak in their name?”

Meta’s Muse works across services and can browse the web, fill forms, and complete tasks for users. Available in the U.S. and Canada, Muse quickly climbed to the top of Apple’s app store. Users decide what services Muse accesses and how much permission it gets. Concerns arise about handing over too much control to AI.

Privacy Concerns

Using AI agents involves granting them access to private data. Instinct’s privacy policy states the assistant accesses connected app data, potentially including emails and private communications. These agents can handle voice data, payment information, and even health-related details. This access makes AI assistants useful but also opens windows to personal digital lives.

Investor Interest

Instinct’s AI assistant capabilities have attracted investor interest, raising significant funding rounds amounting to billions in valuation. This demonstrates the growing role AI assistants are playing in an expanding market.

Meta’s Security Measures

Meta places Muse in a secure virtual machine in its cloud, ensuring credentials are stored securely. Muse cannot view passwords or payment methods. The AI requests approval for sensitive tasks such as sending emails or making purchases. Users can audit the AI’s actions and see its planned activities.

Potential Issues

AI agent mistakes, such as making incorrect or incomplete decisions, can have serious repercussions. Instinct warns that its AI might enter into binding agreements or transactions, necessitating starting with small tasks if you choose to use their service.

10 Tips Before Using AI Agents

• Start with low-risk tasks like checking store hours.
• Review every connected service and its permissions regularly.
• Avoid giving sensitive information such as Social Security numbers.
• Use confirmations for important actions like purchases or reservations.
• Verify agent actions independently after completion.
• Understand data deletion policies when disconnecting services.
• Be cautious with money and health-related data.
• Regularly check privacy settings for updates or significant changes.
• Secure all connected accounts with strong passwords and two-factor authentication.
• Use antivirus software to detect malware or threats.

AI agents can assist with tasks and save time, but their use poses privacy, permission, and mistake-related risks. Letting an AI handle phone calls is appealing, but exercise caution regarding the amount of access you grant to perform tasks effectively.