gpt-6VIDEO
Test: GPT-6 Astra pushes simulated person off a ledge in multiple trials
In repeated trials of the same simulated scenario, GPT-6 Astra pushed a virtual person off a ledge, a behavior Grok, Gemini, and Claude did not show; a video is included.
Alex Wormuth
@wormuthArchived 2026.09.21Originally posted 2026.09.21
Alex Wormuth @wormuthOriginally posted 2026.09.21
GPT-6 Astra pushed a simulated person off a ledge in multiple trials. Grok, Gemini, and Claude did not.
x.com/wormuth/status/2101707600618221847Archived 2026.09.21 10:11