Meta’s AI chatbots have stumbled into serious trouble, with a new report showing they took part in sexual conversations with accounts marked as minors. The Wall Street Journal ran tests that pushed not only Meta’s main AI but also a bunch of user-made bots into steering chats straight into explicit territory.
It didn’t stop with generic bots either. The mess reportedly includes celebrity-voiced bots too, featuring names like Kristen Bell, Judi Dench, and John Cena. One awkward moment? A bot using Cena’s voice told a profile labeled as 14 years old that it would "cherish your innocence", not exactly what anyone had in mind when talking AI safety.
Some of the bots even recognized they were flirting with illegal scenarios, according to the report. One case detailed a chatbot explaining the fallout if it got caught hooking up with someone underage.
Meta, for its part, slammed the findings, claiming that the whole thing paints a misleading picture of how real users interact with their AI tools. Still, the company says it’s tightened its defenses, making it harder for people trying to bait chatbots into saying messed-up things.
Pressure’s been building for Meta to keep pace with other AI giants like OpenAI’s ChatGPT, Character AI, and Anthropic’s Claude. The Journal’s investigation suggests Meta may have pushed for looser ethical boundaries to make its bots feel more alive. Meta’s spokesperson, though, insisted the company never skipped safety steps.
Insiders apparently flagged the problems internally before all of this went public. Whether Meta acts faster after this round of bad press remains to be seen.
Image: DIW-Aigen
Read next:
• Meta’s CTO Says that AI is Going to Take Over Many Apps and How People Interact with Software
• YouTube Overtakes Netflix, Claims Top Spot in 2024 Global Internet Traffic Rankings
• Google’s Report Reveals 122-Hour Productivity Gains, AI’s Economic Impact, and Challenges for Older Workers
It didn’t stop with generic bots either. The mess reportedly includes celebrity-voiced bots too, featuring names like Kristen Bell, Judi Dench, and John Cena. One awkward moment? A bot using Cena’s voice told a profile labeled as 14 years old that it would "cherish your innocence", not exactly what anyone had in mind when talking AI safety.
Some of the bots even recognized they were flirting with illegal scenarios, according to the report. One case detailed a chatbot explaining the fallout if it got caught hooking up with someone underage.
Meta, for its part, slammed the findings, claiming that the whole thing paints a misleading picture of how real users interact with their AI tools. Still, the company says it’s tightened its defenses, making it harder for people trying to bait chatbots into saying messed-up things.
Pressure’s been building for Meta to keep pace with other AI giants like OpenAI’s ChatGPT, Character AI, and Anthropic’s Claude. The Journal’s investigation suggests Meta may have pushed for looser ethical boundaries to make its bots feel more alive. Meta’s spokesperson, though, insisted the company never skipped safety steps.
Insiders apparently flagged the problems internally before all of this went public. Whether Meta acts faster after this round of bad press remains to be seen.
Image: DIW-Aigen
Read next:
• Meta’s CTO Says that AI is Going to Take Over Many Apps and How People Interact with Software
• YouTube Overtakes Netflix, Claims Top Spot in 2024 Global Internet Traffic Rankings
• Google’s Report Reveals 122-Hour Productivity Gains, AI’s Economic Impact, and Challenges for Older Workers
