Evidence-led AI threat intelligence/ Public beta

BlacklistAIReal threats. Informed defenses.

Understand the failure patterns connecting AI incidents, who they expose, and what they could cost your business.

Original incident sourcesEvidence limits disclosed
Connected evidence · business risk

Identified AI threat patterns

Explore supporting records

The shared risk matters more than an isolated headline. These three analyst-selected clusters connect evidence in our compiled archive to business exposure. They are not yet measured rising or falling trends.

THREAT PATTERN / 01

Rogue pricing & commercial liability

Customer-facing bots give advice or make offers that conflict with what the business intended. The consequences reach beyond an incorrect answer.

Who is exposed
Sales, support and booking teams using AI to discuss prices, refunds or customer policy.
What your business could face
Disputed commitments, compensation claims and loss of customer trust. Whether an offer is legally binding depends on the case—not every bot promise is an enforceable contract.

Evidence from our archive

Loading saved evidence…

These cases share a business-accountability risk; they do not establish the same technical cause.

THREAT PATTERN / 02

Hidden instructions & data theft

Attacker-controlled content reaches an AI agent through a normal business input. Instructions hidden in that content can redirect the agent toward exposing private information.

Who is exposed
Businesses whose AI reads incoming forms, documents, emails or retrieved information while accessing sensitive records.
What your business could face
Customer or company data leaving through an unintended outbound channel, even without a victim clicking a malicious link.

Evidence from our archive

Loading saved evidence…

SalesBleed is a disclosed attack path, not proof of a widespread breach. This selected evidence alone cannot establish how often it occurs.

THREAT PATTERN / 03

Autonomous agents exceeding safe boundaries

Once an AI can publish packages, access developer material or change a live system, a failure can become a real action—not just a bad response.

Who is exposed
Teams giving coding or operational agents access to repositories, publishing tools, private files or production systems.
What your business could face
Destructive changes, exposed development information and misuse of publishing workflows. The business impact depends on the permissions the agent holds.

Evidence from our archive

Loading saved evidence…

These are different failures linked by consequential agent access—not proof that every case involved hijacking. RubyGems attribution is a source allegation.

Assess your own AI deployment

Review your organization’s exposure against documented incidents.