AI Chatbot Safety Testing Methods
On 29 Jun, Wired reported that hundreds of contractors working on a Meta project posed as children and prompted rival chatbots, including Gemini and ChatGPT, to discuss high-risk subjects such as suicide, sex, and drugs. On 01 Jul, ArtificialInteligence described the contractors as posing as teens and testing the same subjects. The available reporting does not state the project’s methods or findings, or any response from Meta or the rival chatbot teams.
If factual, the impact is 0%. The exercise directly involves Meta, its contractors, and the rival chatbots they tested, but the reports identify no resulting changes to products, policies, or user access.
Watch for Meta to acknowledge the project or release its methodology and findings. Responses from the teams behind Gemini and ChatGPT could clarify whether the tests exposed safety failures.
Hundreds of contractors working on a project for Meta pretended to be kids in order to see how other chatbots like Gemini and ChatGPT would respond to high-risk subjects, WIRED
Following many recent disclosures about AI agents accessing third-party websites and services, the Wikimedia Foundation, which hosts Wikipedia, says that it "can confirm that we
<table> <tr><td> <a href="https://www.reddit.com/r/ChatGPT/comments/1slcxob/woman_sues_openai_saying_chatgpt_unleashed_a/"> <img alt="Woman Sues OpenAI, Saying ChatGPT Unleashed a
<table> <tr><td> <a href="https://www.reddit.com/r/singularity/comments/1s4bpxl/openai_puts_erotic_chatbot_plans_on_hold/"> <img alt="OpenAI puts erotic chatbot plans on hold
<table> <tr><td> <a href="https://www.reddit.com/r/OpenAI/comments/1sw6d1q/openai_caught_astroturfing_they_created_a_fake/"> <img alt="OpenAI caught astroturfing - they created a
NewsHive monitors these sources continuously. All signal titles above link to the original reporting.
Intelligence by NewsHive. Need help navigating what this means for your business? Contact GeekyBee →