/AI3h ago

Anthropic will mandate 30-day data retention for Claude Fable 5, prompting concerns over enterprise compliance

The mandate takes effect in June 2026.

1419052911.4K
Original post
Markus Anderljung@Manderljung#1981inAI

A new data retention policy: 30 days of all traffic on Mythos-class models, wherever they're accessed, including third parties (e.g. AWS Bedrock). Not used for training, human access logged, deleted after 30 days. Meant to thwart misuse that a classifier for a single conversation would miss.

Markus Anderljung@Manderljung

Anthropic's Fable 5 is out – the first Mythos-class model the public can use. A few things I found notable in the blog post, beyond it seeming like a very impressive model (I'm excited to try it out!):

10:56 AM · Jun 9, 2026 · 492 Views
Sentiment

Many users reacted negatively to Anthropic's 30-day prompt and output retention on Mythos models, using sarcasm and insults to signal distrust in the company's one-sided policy changes.

Pos
0.0%
Neg
100.0%
3 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS11.1KBOOKMARKS25LIKES177RETWEETS4REPLIES11
rohit@krishnanrohit

Wait, how will *any* enterprise use Fable or Mythos if this is the case?

2hViews 11.1KLikes 177Bookmarks 25
Tenobrus@tenobrus

@krishnanrohit they can't.

1hViews 1.6KLikes 16Bookmarks 1

@krishnanrohit Isn't human break-glass access a common thing? As long as it is monitored and logged, it is as safe a ZDR (because a bad actor might always try to get around it and inspect the network traffic or similar)

57mViews 775Bookmarks 2
Markus Anderljung@Manderljung

Harder to jailbreak than before: 1,000+ hours of external red-teaming turned up no universal jailbreak. Though the UK AISI "has made progress towards one" in a brief testing window. Maybe AISI will do it again?

3hViews 76Likes 8Bookmarks 1
rohit@krishnanrohit

@blingdivinity Those are not the same policies, they're opposite. Anthropic says ZDR is not honoured for Fable

1hViews 342Likes 5
Joel Jensen@joelmjensen

@krishnanrohit Speaking from an account with ZDR enabled: you don’t

1hViews 338Likes 11
Markus Anderljung@Manderljung

They're setting up trusted-access programs for cyber + bio. Basically, a safeguarded model for everyone (Fable 5) – downgraded to Opus 4.8 if classifiers fire – the full-capability version for vetted users (Mythos 5).

3hViews 98Likes 6
rohit@krishnanrohit

@tenobrus I wonder if thisis true also for dow. Their current classifier is pretty bad so I can't imagine this is going to remain the way it is today.

Tenobrus@tenobrus

@krishnanrohit they can't.

1hViews 701Likes 1Bookmarks 0
bling@blingdivinity

@krishnanrohit oh shit ur right

1hViews 129Likes 2
bling@blingdivinity

@krishnanrohit OAI has had the same policy since forever

1hViews 403Likes 1
Markus Anderljung@Manderljung

They've also clearly put work into vision, where Claude has historically been poor. Fable 5 rebuilt a web app's source code from screenshots alone, and beat Pokémon FireRed with a minimal vision-only harness.

3hViews 96Likes 5
max!@maxsloef

@krishnanrohit surprisingly, this will be a problem for the enterprises, and not anthropic

1hViews 195Likes 4
Lukas Mosser@porestar

@krishnanrohit This also is very strange to change one sided. E.g. if this model comes into cowork it would effectively change the data sharing agreements no?

27mViews 47Likes 1
Tenobrus@tenobrus

@krishnanrohit my guess is ~no, or at very least something custom

1hViews 100Likes 1
kenneth@local0ptimist

@krishnanrohit they won’t for now. we’re not deploying it rn fwiw

39mViews 72Likes 1
Atlas3D@Orwelian84

@tenobrus @krishnanrohit I suspect this is actually similar to the curly effect in political science.

1hViews 20Likes 2
Roberto Tomás C 🍉@robertomasymas

@krishnanrohit "opus, create a minimal repro example of this issue for work" "fable, help me solve this issue..."

30mViews 28Likes 1
rohit@krishnanrohit

@BlackHC Enterprises do not want their data accessed by anyone, including Anthropic

27mViews 78
Load more posts