Logiciel Contact Us
Success Stories Tech News Contact Us
whitepaper

The AI Attack Surface Your Application Security Scanner Cannot See.

Static analysis reads code, dependency scanners read manifests, and neither reads the retrieved document that tells your agent to email the customer list outside the company. This report sets out the five failure classes that cover almost everything exploited in production, and the four design decisions that survive a red team.

In depth

The Exploit Arrives As Content. Your Scanners Stay Quiet.

01

The trap most engineering teams walked into: ship the AI feature against a demo date, wire a broad service account into retrieval, hand the agent one shared credential with write access it never needs, and put a prompt-injection classifier at the input as though there were a grammar of malicious instructions to match.

02

What the teams surviving a red team do: assume the model will be persuaded and bound what it can reach: a privileged planner that never sees untrusted content, one credential per tool scoped to a single action, retrieval authorisation enforced in the query, and schema validation at every consuming boundary.

The detail

What Separates An Architecture That Holds.

Zone · 01

Trust Zones

Not Filters

Split the pipeline. A privileged planner that never sees untrusted content, and an unprivileged worker that reads it but cannot act. Input filtering is not a control here: there is no grammar of malicious instructions to match, and every blocklist is one paraphrase from bypass. Treat classifier detection as telemetry.

Zone · 02

One Credential Per Action

One credential per tool, scoped to the narrowest action the use case needs, so a read agent cannot write and a support agent cannot touch billing. Then put a human in the loop on anything irreversible: payments, deletions, external messages, code merges, permission changes. Test the boundary by attempting what should be impossible.

Zone · 03

Authorization Inside The Query

An index built once by a service account that reads everything will return the board pack to a contractor who asks the right question. Enforce permissions as a pre-filter on the candidate set, because filtering unauthorized documents out after retrieval is a leak waiting for a ranking bug. Add per-role tests on every corpus change.

By the numbers

The figures that make it a board-level conversation.

13%
of organisations reported a breach involving an AI model or application, and 97% of those had no AI-specific access controls in place
$670K
added to the average breach cost where shadow AI featured. 20% of organisations traced a breach to unsanctioned AI use
63%
of breached organisations had no AI governance policy at all, and most could not produce an inventory of the AI systems already running in their estate
Inside the report

What you'll take away.

01

Inventory every AI system, including the unapproved ones

Model, version, data sources, tools, credentials and owner for each one. Shadow AI costs roughly $670,000 more per breach precisely because it is invisible, and nobody can scope an incident on a system that appears in no register.

02

Scope every tool to one action

No shared keys and no write tokens where read suffices. Then treat every completion as hostile input to whatever runs next: schema-validate on the way out, parameterise queries, encode contextually and allowlist egress, exactly as you would for public internet traffic.

03

Cap the spend and the depth

Token flooding, adversarially expensive inputs, recursive agent loops and retry storms against a metered upstream all convert traffic into invoice. Per-tenant budget ceilings, a concurrency cap, a maximum tool-call depth and a breaker that fails closed with a defined fallback.

04

Run adversarial evaluation in CI

A versioned suite of injection, leakage and exfiltration cases written against your specific tools and permissions, failing the build when it regresses. Add an external OWASP-aligned exercise before launch, and keep a model bill of materials with pinned digests rather than tags.

Questions

Frequently asked.

Can a prompt-injection classifier not simply block these attacks at the input?

Classifiers cannot do that job. There is no grammar of malicious instructions to match, and payloads can be encoded, translated, split across documents or hidden in white text. Treat them as useful telemetry and a reasonable speed bump. Deploy one as a boundary and a red team will get through it in an afternoon.

Our AI features are internal only, so does the attack surface really apply to us?

Indirect injection does not need your interface. Any content the model ingests is a delivery path: an inbound email, a supplier PDF, a customer support ticket, a scraped page in the corpus. Internal deployment removes the login page, not the ingestion path.

Does making retrieval authorisation-aware mean rebuilding our vector index?

Usually not the embeddings, but yes to the metadata and the query path. You need source-system permission identifiers stored alongside each chunk and applied as a pre-filter before ranking. Budget a sprint for the plumbing, plus per-role tests that keep it honest afterwards.

We already run penetration tests, so why is a separate AI exercise necessary?

Standard scoping tests your code and infrastructure. It does not attempt indirect injection through the corpus, exfiltration through tool chains, embedding inversion or cost exhaustion. Those need adversarial cases written
against your specific tools and permissions, then kept in CI as the system changes.

How much of this can we defer until after the feature has real users?

Tool scoping and output validation cannot be deferred cheaply, because both are wired through the call path. Inventory, logging classification and CI evaluation can follow within a quarter. The pattern in IBM's data is unpriced
deferral: 97% of breached AI systems had no AI-specific access controls.

Get the whitepaper

Have it emailed to you.

Drop your details and we'll send The AI Attack Surface Your Application Security Scanner Cannot See straight to your inbox - no spam, unsubscribe anytime.

Download whitepaper
Next step

Have someone attack it first.

In a two-week trial sprint we deliver a threat model for one live AI feature and an adversarial evaluation suite running in your CI.

Book a security review