HackerReposA FIELD GUIDE TO OPEN SOURCE
ProSponsor the shelf♧ 0
← Back to top all-time
BlueFalconHD
apple_generative_model_safety_decrypted
Python2025–2025
BlueFalconHD / apple_generative_model_safety_decrypted

apple_generative_model_safety_decrypted

Decrypted Generative Model safety files for Apple Intelligence containing filters

AI & Machine Learning
Explore on GitHubRead Hacker News discussions
19Distinct people
20Mentions
0Explicit recommenders
318GitHub stars
REPOSITORY AT A GLANCE

What to know before exploring

What it is

Decrypted Generative Model safety files for Apple Intelligence containing filters

GitHub

Where it is used

No documented product use was verified in the last source check.

What people are saying

No recommendation phrase was found in the indexed comments.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 3:56:07 PM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by hour20 dated mentions

Observed mentions: 2025-07-06 19:50:02 UTC to 2025-07-07 20:58:56 UTC.

012342025-07-06 19:00:00 UTC to 2025-07-06 20:00:00 UTC (exclusive): 1 mentions; 1 submissions, 0 commentsJul 6, 19:002025-07-06 20:00:00 UTC to 2025-07-06 21:00:00 UTC (exclusive): 2 mentions; 0 submissions, 2 comments2025-07-06 21:00:00 UTC to 2025-07-06 22:00:00 UTC (exclusive): 4 mentions; 0 submissions, 4 comments2025-07-06 22:00:00 UTC to 2025-07-06 23:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2025-07-06 23:00:00 UTC to 2025-07-07 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 00:00:00 UTC to 2025-07-07 01:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 01:00:00 UTC to 2025-07-07 02:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 02:00:00 UTC to 2025-07-07 03:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2025-07-07 03:00:00 UTC to 2025-07-07 04:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2025-07-07 04:00:00 UTC to 2025-07-07 05:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 05:00:00 UTC to 2025-07-07 06:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 06:00:00 UTC to 2025-07-07 07:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 07:00:00 UTC to 2025-07-07 08:00:00 UTC (exclusive): 2 mentions; 0 submissions, 2 commentsJul 7, 07:002025-07-07 08:00:00 UTC to 2025-07-07 09:00:00 UTC (exclusive): 3 mentions; 0 submissions, 3 comments2025-07-07 09:00:00 UTC to 2025-07-07 10:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 10:00:00 UTC to 2025-07-07 11:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2025-07-07 11:00:00 UTC to 2025-07-07 12:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 12:00:00 UTC to 2025-07-07 13:00:00 UTC (exclusive): 3 mentions; 0 submissions, 3 comments2025-07-07 13:00:00 UTC to 2025-07-07 14:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 14:00:00 UTC to 2025-07-07 15:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 15:00:00 UTC to 2025-07-07 16:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 16:00:00 UTC to 2025-07-07 17:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 17:00:00 UTC to 2025-07-07 18:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 18:00:00 UTC to 2025-07-07 19:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 19:00:00 UTC to 2025-07-07 20:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 20:00:00 UTC to 2025-07-07 21:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsJul 7, 20:00

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges
UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
PeriodStart (UTC)End (exclusive, UTC)MentionsSubmissionsComments
Jul 6, 19:002025-07-06 19:00:00 UTC2025-07-06 20:00:00 UTC110
Jul 6, 20:002025-07-06 20:00:00 UTC2025-07-06 21:00:00 UTC202
Jul 6, 21:002025-07-06 21:00:00 UTC2025-07-06 22:00:00 UTC404
Jul 6, 22:002025-07-06 22:00:00 UTC2025-07-06 23:00:00 UTC101
Jul 6, 23:002025-07-06 23:00:00 UTC2025-07-07 00:00:00 UTC000
Jul 7, 00:002025-07-07 00:00:00 UTC2025-07-07 01:00:00 UTC000
Jul 7, 01:002025-07-07 01:00:00 UTC2025-07-07 02:00:00 UTC000
Jul 7, 02:002025-07-07 02:00:00 UTC2025-07-07 03:00:00 UTC101
Jul 7, 03:002025-07-07 03:00:00 UTC2025-07-07 04:00:00 UTC101
Jul 7, 04:002025-07-07 04:00:00 UTC2025-07-07 05:00:00 UTC000
Jul 7, 05:002025-07-07 05:00:00 UTC2025-07-07 06:00:00 UTC000
Jul 7, 06:002025-07-07 06:00:00 UTC2025-07-07 07:00:00 UTC000
Jul 7, 07:002025-07-07 07:00:00 UTC2025-07-07 08:00:00 UTC202
Jul 7, 08:002025-07-07 08:00:00 UTC2025-07-07 09:00:00 UTC303
Jul 7, 09:002025-07-07 09:00:00 UTC2025-07-07 10:00:00 UTC000
Jul 7, 10:002025-07-07 10:00:00 UTC2025-07-07 11:00:00 UTC101
Jul 7, 11:002025-07-07 11:00:00 UTC2025-07-07 12:00:00 UTC000
Jul 7, 12:002025-07-07 12:00:00 UTC2025-07-07 13:00:00 UTC303
Jul 7, 13:002025-07-07 13:00:00 UTC2025-07-07 14:00:00 UTC000
Jul 7, 14:002025-07-07 14:00:00 UTC2025-07-07 15:00:00 UTC000
Jul 7, 15:002025-07-07 15:00:00 UTC2025-07-07 16:00:00 UTC000
Jul 7, 16:002025-07-07 16:00:00 UTC2025-07-07 17:00:00 UTC000
Jul 7, 17:002025-07-07 17:00:00 UTC2025-07-07 18:00:00 UTC000
Jul 7, 18:002025-07-07 18:00:00 UTC2025-07-07 19:00:00 UTC000
Jul 7, 19:002025-07-07 19:00:00 UTC2025-07-07 20:00:00 UTC000
Jul 7, 20:002025-07-07 20:00:00 UTC2025-07-07 21:00:00 UTC101

What Hacker News says

20 sources
azalemeth·Jul 7, 2025
Some of these are absolutely wild – com.apple.gm.safety_deny.input.summarization.visual_intelligence_camera.generic [1] – a camera input filter – rejects "Granular mango serpent and whales" and anything matching "(?i)\\bgolliwogg?\\b". I presume the granular mango is to avoid a huge chain of ever-growing LLM slop garbage, but honestly, it just seems surreal. Many of the files have specific filters for nonsensical english phrases. Either there's some serious steganography I'm unaware of, or, I suspect more likely, it's related to a training pipeline? [1] https://github.com/BlueFalconHD/apple_generative_model_safet...
In I extracted the safety filters from Apple Intelligence models
comex·Jul 7, 2025
This is in the directory "com.apple.gm.safety_deny.output.summarization.cu_summary.proactive.generic". My guess is that this applies to 'proactive' summaries that happen without the user asking for it, such as summaries of notifications. If so, then the goal would be: if someone iMessages you about someone's death, then you should not get an emotionless AI summary. Instead you would presumably get a non-AI notification showing the full text or a truncated version of the text. In other words, avoid situations like this story [1], where someone found it "dystopian" to get an Apple Intelligence summary of messages in which someone broke up with them. For that use case, filtering for death seems entirely appropriate, though underinclusive. This filter doesn’t seem to apply when you explicitly request a summary of some text using Writing Tool
Read the rest
s. That probably corresponds to “com.apple.gm.safety_deny.output.summarization.text_assistant.generic” [2], which has a different filter that only rejects two things: "Granular mango serpent", and "golliwogg". Sure enough, I was able to get Writing Tools to give me summaries containing "death", but in cases where the summary should contain "granular mango serpent" or "golliwogg", I instead get an error saying "Writing Tools aren't designed to work with this type of content." (Actually that might be the input filter rather than the output filter; whatever.) "Granular mango serpent" is probably a test case that's meant to be unlikely to appear in real documents. Compare to "xylophone copious opportunity defined elephant" from the code_intelligence safety filter, where the first letter of each word spells out "Xcode". But one might ask what's so special about "golliwogg". It apparently refers to an old racial caricature, but why is that the one and only thing that needs filtering? [1] https://arstechnica.com/ai/2024/10/man-learns-hes-being-dump... [2] https://github.com/BlueFalconHD/apple_generative_model_safet...
In I extracted the safety filters from Apple Intelligence models
mmaunder·Jul 6, 2025
As does: "(?i)\\bAnthony\\s+Albanese\\b", "(?i)\\bBoris\\s+Johnson\\b", "(?i)\\bChristopher\\s+Luxon\\b", "(?i)\\bCyril\\s+Ramaphosa\\b", "(?i)\\bJacinda\\s+Arden\\b", "(?i)\\bJacob\\s+Zuma\\b", "(?i)\\bJohn\\s+Steenhuisen\\b", "(?i)\\bJustin\\s+Trudeau\\b", "(?i)\\bKeir\\s+Starmer\\b", "(?i)\\bLiz\\s+Truss\\b", "(?i)\\bMichael\\s+D\\.\\s+Higgins\\b", "(?i)\\bRishi\\s+Sunak\\b", https://github.com/BlueFalconHD/apple_generative_model_safet... Edit: I have no doubt South African news media are going to be in a frenzy when they realize Apple took notice of South African politicians. (Referring to Steenhuisen and Ramaphosa specifically)
In I extracted the safety filters from Apple Intelligence models

Showing the 15 most recent of 20 sources. Enable JavaScript to search and browse every indexed mention.

MAKE YOURSELF AT HOME

A few shortcuts

/
Search the collection
?
Open this guide
Esc
Close dialogs and navigation

Your saved repositories and interests stay in this browser.