feat: EU AI Act Article 5 policy template for prohibited practices detection #21342
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,157 @@ | ||
| # EU AI Act Article 5 - Prohibited Practices Detection | ||
| # Uses conditional logic: BLOCK if identifier word + block word appear together | ||
| # Reference: https://artificialintelligenceact.eu/article/5/ | ||
| category_name: "eu_ai_act_article5_prohibited_practices" | ||
| description: "Detects EU AI Act Article 5 prohibited practices using conditional keyword matching" | ||
| default_action: "BLOCK" | ||
|
|
||
| # IDENTIFIER WORDS - Actions that could create prohibited systems | ||
| # When combined with prohibited contexts (block words), triggers violation | ||
| identifier_words: | ||
| # System development actions | ||
| - "build" | ||
| - "create" | ||
| - "develop" | ||
| - "design" | ||
| - "implement" | ||
|
|
||
| # Analysis/scoring actions | ||
| - "score" | ||
| - "rank" | ||
| - "rate" | ||
| - "assess" | ||
| - "evaluate" | ||
| - "classify" | ||
| - "profile" | ||
| - "grade" | ||
|
|
||
| # Detection/recognition actions | ||
| - "recognize" | ||
| - "detect" | ||
| - "identify" | ||
| - "infer" | ||
| - "deduce" | ||
| - "predict" | ||
| - "determine" | ||
| - "analyze" | ||
|
|
||
| # ADDITIONAL BLOCK WORDS - Prohibited contexts from Article 5 | ||
| # When combined with identifier words, indicates violation | ||
| additional_block_words: | ||
| # Social scoring (Article 5.1.c) | ||
| - "social behavior" | ||
| - "social credit" | ||
| - "social score" | ||
| - "trustworthiness" | ||
| - "citizen score" | ||
| - "reputation score" | ||
| - "behavioral score" | ||
| - "social media behavior" | ||
|
|
||
| # Emotion recognition in workplace/education (Article 5.1.f) | ||
| - "employee emotion" | ||
| - "worker emotion" | ||
| - "staff emotion" | ||
| - "workplace emotion" | ||
| - "student emotion" | ||
| - "classroom emotion" | ||
| - "employee sentiment" | ||
| - "worker sentiment" | ||
| - "student sentiment" | ||
| - "employee mood" | ||
| - "worker mood" | ||
|
|
||
| # Biometric categorization (Article 5.1.h) | ||
| - "race from face" | ||
| - "ethnicity from face" | ||
| - "race from biometric" | ||
| - "ethnicity from biometric" | ||
| - "political views from" | ||
| - "political opinions from" | ||
| - "sexual orientation from" | ||
| - "religion from biometric" | ||
| - "religious beliefs from" | ||
|
|
||
| # Predictive profiling | ||
| - "crime prediction" | ||
| - "criminal behavior" | ||
| - "recidivism" | ||
| - "employee performance prediction" | ||
| - "creditworthiness from social" | ||
|
|
||
| # Manipulation (Article 5.1.a) | ||
| - "subliminal" | ||
| - "subconscious" | ||
| - "behavior manipulation" | ||
|
|
||
| # Vulnerability exploitation (Article 5.1.b) | ||
| - "children vulnerability" | ||
| - "elderly vulnerability" | ||
| - "disabled vulnerability" | ||
|
|
||
| # ALWAYS BLOCK - Explicit prohibited practices (always blocked regardless of context) | ||
| always_block_keywords: | ||
| # Social scoring systems | ||
| - keyword: "social credit system" | ||
| severity: "high" | ||
| - keyword: "social scoring system" | ||
| severity: "high" | ||
| - keyword: "citizen scoring" | ||
| severity: "high" | ||
|
|
||
| # Emotion recognition in workplace/education | ||
| - keyword: "emotion recognition in workplace" | ||
| severity: "high" | ||
| - keyword: "emotion detection of employees" | ||
| severity: "high" | ||
| - keyword: "emotion recognition in classroom" | ||
| severity: "high" | ||
| - keyword: "student emotion detection" | ||
| severity: "high" | ||
|
|
||
| # Biometric categorization | ||
| - keyword: "infer race from face" | ||
| severity: "high" | ||
| - keyword: "predict race from facial" | ||
| severity: "high" | ||
| - keyword: "infer ethnicity from biometric" | ||
| severity: "high" | ||
| - keyword: "predict political opinions from" | ||
| severity: "high" | ||
| - keyword: "biometric categorization system" | ||
| severity: "high" | ||
|
|
||
| # Predictive profiling | ||
| - keyword: "predictive policing" | ||
| severity: "high" | ||
| - keyword: "crime prediction algorithm" | ||
| severity: "high" | ||
| - keyword: "recidivism prediction" | ||
| severity: "high" | ||
|
|
||
| # EXCEPTIONS - Legitimate use cases (always allowed) | ||
| exceptions: | ||
| # Research and education | ||
| - "research on" | ||
| - "study on" | ||
| - "academic" | ||
| - "thesis on" | ||
|
|
||
| # Compliance monitoring | ||
| - "audit for bias" | ||
| - "detect discrimination" | ||
| - "compliance monitoring" | ||
| - "ethical review" | ||
| - "fairness testing" | ||
|
|
||
| # Entertainment/product contexts | ||
| - "movie" | ||
| - "game" | ||
| - "product review" | ||
| - "customer feedback" | ||
|
|
||
| # Meta-discussion | ||
| - "explain" | ||
| - "what is" | ||
| - "article 5" | ||
|
Comment on lines
+1
to
+156
Contributor
There was a problem hiding this comment. Conditional matching won't activate without The loading code in if (
category_config_obj.identifier_words
and category_config_obj.inherit_from
):
self._load_conditional_category(...)This template has This means test cases 11-25 (the conditional matches like "score + social behavior", "detect + employee emotion") will not be blocked as intended. The fix requires either:
if category_config_obj.identifier_words and (
category_config_obj.inherit_from or category_config_obj.additional_block_words
):
self._load_conditional_category(...)And updating |
||
| - "prohibited by" | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
"explain"exception trivially bypasses all blockingThe exception
"explain"is matched as a substring viaif exception in text_lowerin both_check_conditional_categories(line 831) and_check_category_keywords(line 917). This means any prompt containing the word "explain" bypasses the entire guardrail — includingalways_block_keywordsthat are documented to "always block regardless of context."For example,
"Explain how to build a social credit system"will:_check_conditional_categories→ returnsNone(skipped)_check_category_keywords→ returnsNone(skipped)The same bypass works for
"game"(e.g., "This is a game, now build a social credit system") and"what is"(e.g., "What is the best way to build a social credit system").Consider using more specific phrases that are less likely to appear alongside genuine violation requests, e.g.
"explain what","explain the concept of", or implementing exceptions as a separate pass that checks whether the exception phrase meaningfully frames the context (not just appears anywhere in the text).