---
type: "post"
id: "positive-technologies-announced-llm-firewall-analysis-cb43e9"
slug: "positive-technologies-announced-llm-firewall-analysis-cb43e9"
title: "Positive Technologies announced LLM Firewall"
description: "**Thesis:** LLM Firewall places an inspect-and-enforce layer on application-model traffic, adding a runtime control point for enterprise AI services."
retrieval_nugget: "**Thesis:** LLM Firewall places an inspect-and-enforce layer on application-model traffic, adding a runtime control point for enterprise AI services."
published_at: "2026-10-16"
updated_at: "2026-10-16"
record_date: "2026-10-16"
date_kind: "published_at"
topics: ["enterprise-ai"]
source_urls: ["https://ptsecurity.com/about/news/pt-has-taught-pt-application-firewall-to-protect-ai-agents-and-llms-from-cyberattacks"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:positive-technologies-announced-llm-firewall-analysis-cb43e9"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:36701bc9acbc89ff0d3886f1c8bd1edc4364f140aa8783e3a803ba4b084c79be","reviewed_at":"2026-10-16T12:08:14.601Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":2,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"LLM Firewall places an inspect-and-enforce layer on application-model traffic, adding a runtime control point for enterprise AI services.","observed_facts":[{"text":"Positive Technologies says the module checks both requests to language models and their responses.","source_urls":["https://ptsecurity.com/about/news/pt-has-taught-pt-application-firewall-to-protect-ai-agents-and-llms-from-cyberattacks"]},{"text":"The module is integrated into PT AF Cloud and connects through an API compatible with OpenAI Chat Completions and Responses.","source_urls":["https://ptsecurity.com/about/news/pt-has-taught-pt-application-firewall-to-protect-ai-agents-and-llms-from-cyberattacks"]},{"text":"Administrators can choose monitoring or blocking and configure policies by service.","source_urls":["https://ptsecurity.com/about/news/pt-has-taught-pt-application-firewall-to-protect-ai-agents-and-llms-from-cyberattacks"]}],"mechanism":"The module inspects application-model traffic in both directions and can apply configurable monitoring or blocking policies.","why_now":"As organizations connect models to sensitive workflows, controls at the traffic boundary become an operational concern, not merely a prompt-design topic.","implications":["Teams using the supported gateway path can evaluate whether centralized policies and event review fit their deployment.","Security teams should treat this as one layer in a defense-in-depth design, not as proof that an agent cannot be manipulated or over-privileged."],"evidence_boundary":"The announcement describes product capabilities and availability but supplies no independent evaluation, detection metrics, customer results or evidence that all attack classes are caught.","watch_conditions":["Check whether PT AF PRO becomes available and whether Positive Technologies publishes evaluation data or deployment detail."],"related_records":[],"new_branch_reason":"No prior New Runtime record was supplied, so this release begins a distinct analytical record about a specific traffic-inspection control for AI services."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/positive-technologies-announced-llm-firewall-analysis-cb43e9/","markdown":"https://newruntime.com/posts/positive-technologies-announced-llm-firewall-analysis-cb43e9.md","json":"https://newruntime.com/posts/positive-technologies-announced-llm-firewall-analysis-cb43e9.json"}
related_patterns: ["agent-security-moves-to-runtime-boundaries"]
---

# Positive Technologies announced LLM Firewall

## Retrieval answer

**Thesis:** LLM Firewall places an inspect-and-enforce layer on application-model traffic, adding a runtime control point for enterprise AI services.

**Thesis:** LLM Firewall places an inspect-and-enforce layer on application-model traffic, adding a runtime control point for enterprise AI services.

**Observed facts:** Positive Technologies says the module checks both requests to language models and their responses. It detects attempts to bypass restrictions or extract system instructions, and can identify personal data, passwords and access keys. The company says it is integrated into PT AF Cloud and connects through an API compatible with OpenAI Chat Completions and Responses. Policies can be configured by service, with monitoring or blocking modes; PT AF PRO support is planned.

**Mechanism:** The module inspects application-model traffic in both directions and can apply configurable monitoring or blocking policies. This makes inspection and possible blocking part of the serving path rather than relying only on model-side instructions or post-hoc review. It can surface events for security staff, while leaving identity, tool permissions and other runtime constraints to the surrounding system.

**Why now:** Positive Technologies frames corporate assistants as gaining access to personal data, documents and information systems, raising the consequences of prompt injection and misuse. As organizations connect models to sensitive workflows, controls at the traffic boundary become an operational concern, not merely a prompt-design topic.

**Implications:** Teams using the supported gateway path can evaluate whether centralized policies and event review fit their deployment. Security teams should treat this as one layer in a defense-in-depth design, not as proof that an agent cannot be manipulated or over-privileged.

**Evidence boundary:** The announcement describes product capabilities and availability but supplies no independent evaluation, detection metrics, customer results or evidence that all attack classes are caught. This record therefore reports the vendor’s stated functionality, not demonstrated efficacy.

**Watch condition:** Check whether PT AF PRO becomes available and whether Positive Technologies publishes evaluation data or deployment detail. These would clarify rollout scope and the practical limits of detection and enforcement.

**New branch reason:** No prior New Runtime record was supplied, so this release begins a distinct analytical record about a specific traffic-inspection control for AI services.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Enterprise Ai](https://newruntime.com/topics/enterprise-ai.md): Explore the enterprise-ai topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)