Anthropic launches Enterprise Frontier Safeguards, which let businesses control how their data is reviewed, stored, and managed, after pushback from customers
Anthropic on Tuesday announced it will walk back a controversial data retention policy after “a lot of feedback” from its business customers …
The move fits Anthropic’s earlier revision of its Responsible Scaling Policy, which separated commitments it makes itself from recommendations for the industry. The company is pairing greater customer control with safeguards intended to detect misuse of frontier models.
First-order effects
Enterprise customers gain greater control over whether and how their data is reviewed, stored, and managed, reducing the operational objection raised by Anthropic’s prior retention approach.
Anthropic replaces a contested one-size-fits-all retention policy with Enterprise Frontier Safeguards, making privacy controls part of its frontier-model access terms.
Second-order effects
Customers that require data to remain in their own cloud environments can assess Anthropic’s frontier offerings without accepting provider-held retention as the default.
Anthropic’s safety teams must demonstrate that misuse detection can operate alongside the privacy posture promised to enterprise customers, rather than relying on broad access to customer data.
Third-order effects
If enterprise buyers keep treating control of data location and review as a procurement requirement, frontier-model vendors will increasingly compete on deployable safety controls as well as model capability.
Frontier-model governance is likely to shift toward negotiated, customer-specific access conditions, with privacy architecture becoming a security boundary rather than a separate compliance feature.
The trend: Frontier AI providers are moving from uniform data-retention rules toward privacy-preserving, managed access controls for enterprise deployments.
We're also introducing Enterprise Frontier Safeguards (EFS), which give enterprise customers complete privacy (the same as zero data retention), while still being state-of-the-art at preventing adversarial use. EFS rolls out in phases, starting this fall. https://www.anthropic.co…
“Today we're announcing Enterprise Frontier Safeguards (EFS), a solution that combines the privacy of zero data retention (ZDR) with state-of-the-art safeguards for detecting misuse. …
Today we announced our Fable 5.1 model alongside Enterprise Frontier Safeguards, which lets customers deploy this class of frontier model while keeping data in their cloud infrastructure. …
Excited to finally talk about Enterprise Frontier Safeguards, which we announced today with Fable 5.1, our strongest model yet for coding and knowledge work. …
Interesting compromise from Anthropic on ZDR for Fable/Mythos 5.1. Data gets retained, but customers have control of where it's hosted, and when human review is needed they can conduct the review themselves. www.anthropic.com/news/enterpr... [image]
Finally, we've improved our safeguards. Our cybersecurity safeguards now flag benign requests about 60% less often. On basic biology and medical questions, we've recently reduced the fallback rate by around 85%.
Anthropic has found a way to balance their safety and ZDR; I expect this to be very good for enterprise and will serve as a good test for whether Fable usage being at 11% was merely because of ZDR.
Interesting compromise on ZDR. Data gets retained, but customers have control of where it's hosted, and when human review is needed they can conduct the review themselves.
Companies care about privacy b/c they don't want to leak sensitive data, violate data protection laws or legal privilege, or have their data used to train models. Yet malicious actors also use AI for pretty awful things — grooming children, defrauding consumers, planning terroris…
HAHAHAHA it's already happening - “Mythos 5.1 also appears more capable of evading monitors while carrying out a covert side task than all other models tested in some evaluations.” -