Microsoft Research's Dimitris Papailiopoulos finds Claude Opus 4.8 flags a benign algebra problem containing the word 'cybersecurity
The model correctly solved the equation despite the warning.
38419121132.4K
Sentiment
Users criticized Claude AI's safety filters as overly harsh and eroding trust after they triggered banners on simple math queries and the keyword cyber.
Pos
0.0%
Neg
100.0%
5 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS157LIKES5
Daniel Auras@rasdani_
@DimitrisPapail lobotomized beyond recognition
some one should call the model welfare team wait ....
2hViews 157Likes 5
REPLIES1
elie@eliebakouch
@DimitrisPapail if "cyber" in prompt: SAFETY_FILTER = True
Dimitris Papailiopoulos@DimitrisPapail
@eliebakouch LOOK AT THIS
3hViews 73Likes 1Bookmarks 0
John Lussier@John_lussier_
@DimitrisPapail Absolute garbage. @AnthropicAI
2hViews 122Likes 1
Conpertus Workloads@Conpertum
@DimitrisPapail Oh no.
2hViews 68
Roberto Tomás C 🍉@robertomasymas
@DimitrisPapail Really gives you trust that they have a sense of what to filter
2hViews 55
Aaryan Kakad@aaryan_kakad
@DimitrisPapail damn, it is too harsh
2hViews 24
some guy on the internet🧬💻@foldingmachiner
@eliebakouch @DimitrisPapail It's even worse than this. I have that I do biochemistry research in memory and...
2hViews 5