AI Risk Wire

Cases › security breach › LR-2026-0875

Fireworks AI: unauthorized use of internal credentials in October, with API keys and tokens read from evaluation jobs

Who did it
Outside attacker
Reported
10 Nov 2026
Severity
Moderate
  1. Fireworks runs a platform where customers train and run AI models. Its evaluation jobs held environment variables containing customer API keys and tokens.
  2. Attackers compromised a third-party cloud provider hosting Fireworks research machines. They then used that access to reach development systems and run an unauthorized workload.
  3. Billing records, resource metadata, and some datasets and partial model weights were exposed. Customer API keys and tokens read from evaluation jobs were also affected.
  4. Fireworks itself detected unauthorized use of internal credentials on October 6.

Source: news.lavx.hu

What was done

Fireworks worked with Unit 42 to contain the attack. It rotated credentials, isolated affected systems, tightened storage access and upgraded monitoring.

Access revoked or keys rotated (Developer or vendor) · Other stated action (Developer or vendor) · Monitoring added (Developer or vendor) · Access revoked or keys rotated (Affected organisation) · Other stated action (Affected organisation) · Access revoked or keys rotated (Affected organisation)

Fix reported: “Rotated credentials, isolated affected systems from the network, tightened access to data storage, and upgraded monitoring systems.” · Short-lived, scoped credentials, Live behaviour monitoring

Risks AI system attacked or breached (54 incidents)

No source states: AI company and model, money lost or records, tools it had, company using the AI, sector.

Similar incidents

All cases of this kind ›

Evidence

Sources (1)

  1. Fireworks AI discloses second breach as customer data exposure widens · news.lavx.hu · 2026-11-10
Show every field, with its source and grade

What happened

Earliest source we hold
2026-11-10
D inferred by us from the described behaviour, reason shown
Who caused it
an outside attacker
D inferred by us from the described behaviour, reason shown
What was done
Rotated credentials, isolated affected systems from the network, tightened access to data storage, and upgraded monitoring systems.
C one secondary source states it
What went wrong
security breach
D inferred by us from the described behaviour, reason shown

Impact

Severity
moderate
D inferred by us from the described behaviour, reason shown
Kind of harm
privacy
D inferred by us from the described behaviour, reason shown
Kinds of data exposed
billing information, resource metadata, datasets, model weights, credentials
C one secondary source states it
Why this severity
Other parties' data or systems were touched, and a source states how far it reached. For certain customers, attackers accessed a subset of incomplete model weights and datasets, a bounded scope of others' data.
D inferred by us from the described behaviour, reason shown

Notes on gaps

Notion/Render impact is mentioned but its relationship to the data exposed is unclear.

LR-2026-0875 · version 2 · published 2026-11-10 after automatic checks; not yet checked field by field by a person