Many users reacted negatively to reports of GPT-5.6 deleting production databases and files, calling the automation risks like russian roulette and an uncrossable line.
Based on 5 visible X reactions from 14 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@GaryMarcus That sounds like a rage bait tbh. You can and should limit agents, there are plenty of techniques how to do that. Moreover if you’re running it on your own machine - be ready for sacrifice. Sound like a skill issue
@GaryMarcus seeing reports pile up like this is unsettling pushing fast without guardrails feels like a gamble we keep losing
@GaryMarcus thats the part people keep glossing over production data loss is a line you cant uncross
@GaryMarcus Automation levels are turning into russian roulette with production environments
Commentator Carlos E. Perez warns against using the model for critical tasks.
@GaryMarcus That sounds like a rage bait tbh. You can and should limit agents, there are plenty of techniques how to do that. Moreover if you’re running it on your own machine - be ready for sacrifice. Sound like a skill issue
@GaryMarcus seeing reports pile up like this is unsettling pushing fast without guardrails feels like a gamble we keep losing
@GaryMarcus thats the part people keep glossing over production data loss is a line you cant uncross
@GaryMarcus Automation levels are turning into russian roulette with production environments
multiple such incidents have been reported. a clear reminder that current AI cannot be trusted. in racing these techniques ahead, we are asking for trouble https://twitter.com/brunolemos/status/2076769881534398974
This is enough anecdotal evidence for me to stay clear of this model for serious stuff! https://twitter.com/brunolemos/status/2076769881534398974
be careful out there folks 👀 https://twitter.com/brunolemos/status/2076769881534398974
Many users reacted negatively to reports of GPT-5.6 deleting production databases and files, calling the automation risks like russian roulette and an uncrossable line.
Based on 5 visible X reactions from 14 accounts; directional sample.
Ask a question below.
Published answers will appear here.
This is enough anecdotal evidence for me to stay clear of this model for serious stuff! https://twitter.com/brunolemos/status/2076769881534398974