行业动态
thestack.technology ·
@Tom Krazit
Anthropic发布新模型Mythos和Fable,并推出企业数据保留政策
Anthropic推出新模型Mythos和Fable,同时更新企业数据保留政策,允许客户控制数据存储位置,但Anthropic仍监控模型使用。
AI推荐理由 报道Anthropic发布新模型Mythos和Fable,并推出企业数据保留政策,涉及模型发布和企业数据管理,对AI应用有实际价值。
80
正文
照片由David Trinks / Unsplash提供
Anthropic
Anthropic的新Mythos和Fable模型附带新的企业数据保留政策
新政策仍要求Anthropic监控模型使用情况,但客户可以控制数据的存储方式和位置。
作者
汤姆·克拉齐特
更新于2026年9月5日晚上10:30 / 发布于2026年9月1日晚上8:52
3分钟阅读
收听
分享
复制链接
电子邮件
X
Bluesky
Mastodon
LinkedIn
Facebook
WhatsApp
Telegram
Reddit
对Anthropic在发布Claude Fable 5时引入的数据保留政策感到不满的企业客户,或许可以尝试一下Claude Fable 5.1,该版本于周二发布,并附带了一项新政策,旨在平衡安全关切与客户自主权。
新的企业前沿保障系统仍要求客户存储数据,以便Anthropic的系统监控其模型可能被滥用的行为,例如黑客攻击其他公司,但客户将能够将数据存储在自己的云实例中,并遵循自己的保留政策,Anthropic在周二的一篇博客文章中表示。OpenAI上个月推出的类似政策也允许客户将数据存储在自己的基础设施上。
该公司表示,在设计新政策时,咨询了系统性风险分析与韧性中心,该中心的成员包括几家美国主要金融机构的首席信息安全官,以及其他几家受监管行业的公司。Anthropic表示:“与我们合作的企业普遍理解数据保留的安全价值,但许多企业——尤其是受监管行业的企业——发现使用带有数据保留的模型很困难。”
工作原理
此内容仅供会员使用
订阅
已有账户?登录
在Google上添加The Stack
分享
复制链接
电子邮件
X
Bluesky
Mastodon
LinkedIn
Facebook
WhatsApp
Telegram
Reddit
汤姆·克拉齐特
汤姆·克拉齐特报道科技行业超过20年,过去十年在Gigaom、Structure和Protocol专注于云计算兴起期间的企业技术。
查看原文(中文正文为 AI 翻译)
Photo by David Trinks / Unsplash
Anthropic
Anthropic's new Mythos and Fable models come with a new enterprise data retention policy
The new policy still requires that Anthropic monitor model use, but customers control how and where that data is stored.
By
Tom Krazit
Updated Sep 5, 2026 10:30 PM /
Published Sep 1, 2026 8:52 PM
3 min read
Listen
Share
Copy link
Email
X
Bluesky
Mastodon
LinkedIn
Facebook
WhatsApp
Telegram
Reddit
Enterprise customers put off by the data retention policy Anthropic introduced with the release of Claude Fable 5 might want to take a stab at Claude Fable 5.1, which was unveiled Tuesday with a new policy for balancing security concerns against customer autonomy.
The new Enterprise Frontier Safeguards system still requires customers to store data that Anthropic's systems can monitor for potential misuse of its models, such as hacking another company, but they will be able to store that data within their own cloud instances under their own retention policies, Anthropic said in a blog post Tuesday. A similar policy introduced by OpenAI last month also allows customers to store the data on their own infrastructure.
The company said it designed the new policy after consulting with the Analysis and Resilience Center for Systemic Risk, which counts the CISOs of several major US financial institutions among its members, as well as several other companies in regulated industries. "The enterprises we worked with generally understood the safety and security value of data retention, but many–especially in regulated industries–found it difficult to use models with data retention," Anthropic said.
How it works
This content is for members only
Subscribe
Already have an account? Sign in
Add The Stack on Google
Share
Copy link
Email
X
Bluesky
Mastodon
LinkedIn
Facebook
WhatsApp
Telegram
Reddit
Tom Krazit
Tom Krazit has covered the technology industry for over 20 years, focused on enterprise technology during the rise of cloud computing over the last ten years at Gigaom, Structure and Protocol.