S-2026-09-09-orchid-security-agent-drift-kill-switch

Orchid Security — identity drift detection and application-level kill switches for AI agents (September 2026)

Tag: S-2026-09-09-orchid-security-agent-drift-kill-switch Type: article (vendor press release via GlobeNewswire, fetched in full; low-authority relay news4hackers also fetched) Author(s): Orchid Security (New York and London) Date of source: 2026-09-09 Date ingested: 2026-09-10 Authority weight: medium — primary vendor release; self-interested; names regulators/standards but all capability and audit-trail claims are unverified; sole customer voice is non-FS and does not confirm use of the new features Raw file: /_raw_sources/S-2026-09-09-orchid-security-agent-drift-kill-switch.md

What it claims

Orchid Security, an identity-security vendor (“Identity Control Plane”), announced identity drift detection and application-level kill switches for AI agents. Its thesis: agents “do not need to ‘break’ security controls” — they exploit “identity debt” (hard-coded credentials, orphaned accounts, unmanaged authentication paths, excessive permissions) to complete objectives “beyond their initial privilege level within seconds”; its own Identity Gap 2026 research puts 57% of enterprise identity as “unseen and unmanaged”.

The offering is framed as Observe → Understand → Govern → Prove: discover agents and the identities/apps/credentials/tools/access paths they use and capture actual behaviour “not only what was configured in the studio”; compare runtime behaviour with the agent’s original purpose and authorised scope, applying AI-readiness tags and surfacing hygiene gaps; when behaviour or effective authority drifts beyond policy, orchestrate response through existing identity/security/AI infrastructure — reduce permissions, revoke credentials, disconnect tools, suspend workflows, or trigger Orchid’s own application-level kill switch; and generate “a defensible audit trail linking each agent action to the identity used, delegation chain, access path, business context, detected drift and resulting governance response”. It lists five things enterprises “should be able to demonstrate” before scaling agents (identity hygiene, authorization guardrails — “who or what may act, on whose behalf, for what purpose, and under what conditions” — runtime understanding, universal auditability, enforceable response).

The release states these capabilities are now available (following May agentic enhancements), adds integrations with Palo Alto Networks Idira (PAM) and Splunk Enterprise Security, and explicitly cites NIST’s draft Cyber AI Profile and DORA (“an obligation that does not pause because the entity acting is an agent rather than a person”). Orchid will exhibit at the Gartner Security & Risk Management Summit, London, 22–24 Sep.

Notable quotes

• “AI agents do not need to ‘break’ security controls or workflow guardrails. They can find and use the identity debt already embedded across the enterprise” (para 1)
• “In Europe, DORA obliges financial entities to demonstrate control over ICT access and third-party dependencies, an obligation that does not pause because the entity acting is an agent rather than a person.” (Regulators paragraph)
• “Orchid generates a defensible audit trail linking each agent action to the identity used, delegation chain, access path, business context, detected drift and resulting governance response.” (PROVE)
• “AI transformation is exciting. Identity hygiene is not.” — Roy Katmor, CEO

What’s speculative vs. asserted

Asserted (vendor): availability of the five capabilities; the two integrations; the Observe/Understand/Govern/Prove framework; the DORA and NIST references.

Vendor-asserted, unverified: effectiveness of drift detection; the “defensible” audit trail (no retention, immutability, format or export stated); the 57% statistic (own research).

Not stated: how “original purpose and authorized scope” is specified and by whom; any FS customer (the only quote is Findlay Automotive Group, US, and does not confirm use of these features); any EU AI Act or SS1/23 mapping.

Vault inference: mapping of intended-purpose-vs-observed-behaviour drift and delegation-chain audit trails to EU AI Act Art. 12/14 and SS1/23 / SR 11-7 change control; that this is an identity-locus entry into agent governance comparable to Okta Agent SSO / Okta for AI Agents.

Topics this feeds

• AI Governance Platforms — a second IAM vendor (after Okta) claiming agent-governance primitives from the identity layer, and a rare vendor that names DORA explicitly; adds “intended purpose vs observed behaviour” drift as a claimed runtime control.

Open questions raised

• Where is an agent’s “original purpose and authorized scope” recorded, and is that artefact the same as (or reconcilable with) the governance platform’s use-case inventory and the EU AI Act intended-purpose statement?
• Is the audit trail retained, tamper-evident and exportable to a second line or regulator, or a SOC correlation feed (Splunk) only?
• Does an application-level kill switch on an agent operating in production create its own operational-resilience event (DORA incident-classification) — who authorises it and how is the action itself evidenced?