Updated Nov 4
Egeria AI Shield uses separate approvals for WAF blocks and code repair

Egeria AI Shield uses separate approvals for WAF blocks and code repair

PKUTECH describes temporary containment, reviewed repairs and retesting; Sentinel recommendations remain under development.

Containment is not repair

PKUTECH’s [November 4 Japan release](https://pkutech.co.jp/news/2026/11/04/7827/) describes Egeria AI Shield for customer‑owned web apps and APIs. WAF rules provide temporary blocking without changing application code; authorization and business‑logic defects instead need code repair. AI proposes code changes; WAF updates and code deployment have separate human approval and recording steps.

Retest before production

[PKUTECH says](https://pkutech.co.jp/news/2026/11/04/7827/) people approve code differences and the revised app is retested with the same attack scenario before production deployment. Tests run against agreed validation or replica environments. Sentinel currently aggregates records; incident‑specific recommendations are in development. The company calls its October WordPress/PHP exercise a workflow check, not evidence of customer protection. Pair programming, December 2012; archive, not PKUTECH staff or security‑test output. Photo: [Vajrapani666 / Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Pair_Programming_3.jpg), [CC BY‑SA 3.0](https://creativecommons.org/licenses/by‑sa/3.0). Original bytes unchanged; display‑cropped; adaptations retain the same licence.

Tags

AI security

Share this article

PostShare

Related News

AgentCore memory poisoning: what AWS builds in, what its new write-validation pattern adds, and where the demo evidence stops

Nov 11, 2026

AgentCore memory poisoning: what AWS builds in, what its new write-validation pattern adds, and where the demo evidence stops

An AWS Security Blog post from 9 November 2026 separates the memory-poisoning controls AgentCore already offers (IAM namespace conditions, namespace-variable validation, Cedar policy, custom consolidation prompts, the ApplyGuardrail API) from a custom pre-write Lambda interceptor that scores each write and returns ALLOW, QUARANTINE or DENY. The custom layer is a sample you deploy, not a default setting. Its published 11-scenario demo used a keyword-overlap stand-in for embeddings, so the results do not show production accuracy or false-positive rates. Writes that bypass AgentCore Gateway, such as direct SDK calls, are not covered by the interceptor.

AWSAgentCoreAI security
NCSC’s 2028 AI ransomware warning concerns existing weaknesses

Nov 4, 2026

NCSC’s 2028 AI ransomware warning concerns existing weaknesses

NCSC’s forecast for UK critical infrastructure separates future AI-enabled threats from observed incidents and mandatory rules.

AI security
CrowdStrike puts Guardian kill switch at gateway; endpoint follows

Nov 3, 2026

CrowdStrike puts Guardian kill switch at gateway; endpoint follows

The November 3 update separates gateway control, operating-system inspection support and planned Fiddler observability.

CrowdStrikeAI security