OK
Trains on your data by default
= no (Favourable)Only with explicit opt-in.
We do not use your prompts, training data, or API inputs to train or improve our AI models without your explicit opt-in.
service fireworks-aivendor Fireworks AIcategory Inference hosts & routers
Does Fireworks AI train on your data?
No (Favourable)Only with explicit opt-in.
- fields
- 5 favourable1 caution1 unverified1 neutral / n/a
- sources cited
- 5 pages
- source monitor
- 5/5 ok on 2026-10-26
- last field check
Evidence log
OK
Default retention
= none by default (open models) (Favourable)Prompts live only in memory during the request. Exception: the Responses API stores conversations for 30 days because store=True is the default.
Fireworks does not log or store prompt or generation data for any open models, without explicit user opt-in.
Against the 21 tools that state a period (grey dots):
OK
Zero data retention
= default (Favourable)ZDR is the default. Enterprise admins can also enforce a policy that rejects any request that would persist data.
Fireworks has Zero Data Retention by default.
OK
Abuse-monitoring exception
= none stated (no prompt logging for open models) (Favourable)For open models Fireworks logs only metadata such as token counts. No abuse-monitoring retention of prompts is described. Opt-in features and the Responses API (store=true) can keep data.
Fireworks does not log or store prompt or generation data for any open models, without explicit user opt-in.
N/A
Data residency / regions
= US-only option (Enterprise enforcement) (Neutral / n/a)A US endpoint (us.api.fireworks.ai) serves selected models only from the US. Enterprise admins can enforce one region account-wide. Only 'None' and 'US' are listed; other regions via sales. No APAC option documented.
Data residency restricts inference on your account to a single region.
OK
SOC 2 / ISO 27001
= SOC 2 Type II; ISO 27001, 27701, 42001 (Favourable)Listed on Fireworks' compliance docs page; certificates are in the Trust Center.
Fireworks has achieved all three ISO certifications covering security, privacy, and AI management:
UNVERIF
HIPAA BAA
= unverified (Unverified)Fireworks says it 'is HIPAA compliant', but no BAA offer is stated on the pages read.
no official page stated this clearly
WARN
Self-host option
= partial (BYOC, private preview) (Caution)Bring Your Own Cluster runs Fireworks inference in your own Kubernetes cluster, with Fireworks operating the serving stack. Enterprise only, private preview.
Bring Your Own Cluster is in Private Preview for Enterprise customers.
Source checks
- ✓ OKdocs.fireworks.ai/accounts/data-residencyfetched 2026-10-26 · SG/APAC
- ✓ OKdocs.fireworks.ai/ecosystem/integrations/byoc/overviewfetched 2026-10-26 · Self-host
- ✓ OKdocs.fireworks.ai/faq/enterprise/compliance/certificationsfetched 2026-10-26 · SOC2/ISO, HIPAA BAA
- ✓ OKdocs.fireworks.ai/guides/security_compliance/data_handlingfetched 2026-10-26 · Retention, ZDR, Abuse mon.
- ✓ OKfireworks.ai/privacy-policyfetched 2026-10-26 · Trains
Spotted something wrong or out of date? Corrections are welcome; see how to send one. How statuses are assigned.