iCreatAI News
Skip to content
Trending storyUpdating

Baseten publishes security controls for defending against agentic attacks

1 reports1 sources3 hr agoUpdated

Understand this story

AI overview

Baseten has published a security protection system for defending against "agent-speed" cyberattacks: it uses AI-assisted defense, adopts defense in depth with second-level detection and response, and uses projects such as OpenAI Daybreak Red to conduct agentic red-team testing of its own code. Customer-facing controls include: customers can continue to use their own identity systems (IdP) to connect; Baseten API keys are scoped, expire, and can be revoked at any time; deployed model weights can be encrypted with keys from the customer's own cloud KMS; inference runs only in customer-approved regions, and Baseten maintains zero retention of inference data. The above measures are self-published by Baseten, and their actual protective effect has not yet been independently verified.

AI overview based on source reports · Updated 2 hr ago

Coverage timeline

Follow the coverage to explore different perspectives.

Nov 21
  1. Baseten Blog
    Improved controls for the new era of cyberattacks

    Baseten describes a shift toward agent-speed cyberattacks, citing Google's Q4 AI Threat Tracker finding that threat actors are moving from AI-assisted to fully autonomous AI-led attacks, and outlines its defenses including layered controls, AI-assisted detection and response, and agentic red teaming. The post also details customer-facing controls such as IdP-based identity, short-lived tokens, customer-managed KMS keys for weights, network allowlists, region-limited inference with zero data retention, and exportable audit logs. It notes that further posts on API key hygiene, container and network security, identity and access management, and observability are planned.

Activity for this story

Not enough continuous observations to show a trend yet.