AI Risk Wire

Cases › security breach › LR-2026-0529

Mercor data breach via LiteLLM supply-chain compromise (April 2026)

Preliminary Why: Publisher not yet named · subject to change

Developing: a major or severe case captured in the last 72 hours. Later changes are logged on the corrections page.

Status: action reported, no fix

Severity major
Reported 23 Apr 2026

No source states: loss or records, developer and model, agent tool, could do, deployer, industry.

Severity major: test: many third parties affected, as Mercor noted thousands of companies were impacted by the LiteLLM breach

Mercor, a contractor that hires experts to help train AI models, was hit by a security compromise in April 2026. According to Halborn, the attack began with attackers compromising Trivy, a security scanning tool, and then inserting malicious code into LiteLLM, a widely used Python library the company used. Infostealer malware in the LiteLLM code gave the attacker credentials for Mercor systems and environments, and the attacker reportedly stole an estimated 4TB of data including Slack data, ticketing information, videos of contractors, database records and source code. Compromised contractor data includes passport scans, Social Security Numbers and video interviews. The Wall Street Journal reported that at least seven class-action lawsuits were served after the breach, and Meta said it had paused its work with Mercor and was investigating.

What happened

1 · What it reached or didThe attacker reached internal systems and reportedly took an estimated 4TB of data.

What was done

At least seven class-action lawsuits were served after the breach.

Risks

Supply-chain compromise
Primary
24 incidents
Rising: 4 vs 0.2 a week

Read from the sources; not yet reviewed by a person.

Controls

Not yet read for controls. Controls that mitigate the primary risk:

Harness

Sandboxed code execution

Application

Signed and verified components · Vetting of plug-ins, skills and dependencies · AI bill of materials · Vulnerability scanning of AI components · Model provenance checks · Least-privilege access for the agent

Infrastructure / target

Secrets scanning

Similar incidents

All cases of this kind ›

Evidence

No source states: what the AI was doing, what went wrong, how it was caught, when it was found, how many records were exposed, the direct loss, any court case, the cause, what set it off.

Show the full account: 3 items from the primary sources

What happened, in order

  1. 2026-04 Mercor was hacked.
  2. 2026-04-23 WSJ reported at least seven class-action lawsuits served after the breach.

Not dated in the sources:

  • Meta said it had paused its work with Mercor and was investigating.
Show every field, with its source and grade

What happened

When it happened
2026-04-01
C one secondary source states it
Earliest source we hold
2026-04-23
D inferred by us from the described behaviour, reason shown
What went wrong
security breach
C one secondary source states it
primary risk
RK06
D inferred by us from the described behaviour, reason shown
secondary risks
RK04
D inferred by us from the described behaviour, reason shown

Impact

Severity
major
D inferred by us from the described behaviour, reason shown
Kinds of data exposed
passport scans, Social Security Numbers, video interviews
C one secondary source states it
Why this severity
test: many third parties affected, as Mercor noted thousands of companies were impacted by the LiteLLM breach
D inferred by us from the described behaviour, reason shown

Notes on gaps

No specific control is reported as held, failed or missing, so links are empty. Halborn's 4TB figure is an estimate attributed to the attacker and is presented as estimated. The lawsuit allegations are plaintiffs' claims that Mercor disputes, not findings. The claim about unauthorized use of the Mythos model and the Claude Code leak concern other systems and are excluded. The 2026-04-23 timeline date is the WSJ report date, not confirmed as the date of the lawsuits.

LR-2026-0529 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON