Vendor drift
Classified material changes to Wispr AI, Inc.’s public trust documents, from continuous monitoring.
Data as of 2026-09-10 — refreshed on a rolling bake, not live. · wisprflow.ai
Permits training, with an opt-out · verified
as of 2026-08-04
high · Training rights changed · 2026-09-03 · trust_center
Wispr AI's security FAQ now discloses that on Android, when "Improve the model for everyone" is enabled with affirmative consent, both audio recordings and transcripts may be uploaded for model training; uploads are excluded when the user opts out, the org enforces ZDR or disallows data sharing, a HIPAA BAA is active, or policies haven't loaded, and pre-consent recordings aren't uploaded retroactively. This newly documents a training-data processing path for customer dictation content on a trust document covering regulated data.
high · Training rights changed · 2026-08-27 · trust_center
Wispr AI clarified that under Privacy Mode, dictation content is "not kept" on servers rather than merely "not stored in anonymized form," and explicitly states that Privacy Mode is locked on when a HIPAA BAA is active or enterprise enforces Zero Data Retention. The revised language removes ambiguity about training rights and strengthens the commitment that customer content is not used for model training when these protections are enabled, but the change also introduces new complexity around organization-level enforcement that could affect individual user control.
medium · Training rights changed · 2026-08-26 · ai_disclosure
Wispr AI modified its training rights language, shifting from "Privacy Mode" (which explicitly prevents data use for training) to a more general "control whether your data is used to improve or train AI models" framework. The change reframes the commitment from a named, binary privacy feature to a broader opt-out mechanism, potentially weakening the clarity and strength of the training data protection guarantee.
high · AI subprocessor added · 2026-08-24 · dpa
Wispr AI's updated DPA (Aug 24, 2026) adds multiple new AI/ML subprocessors to its processing chain, including Anthropic PBC, Cerebras Systems Inc., OpenAI Inc., OpenPipe Inc., and Fireworks AI, Inc. for natural language processing and AI features. These subprocessors now receive customer audio data and transcription content as part of the Service, representing a material expansion of third-party AI processing of customer data.
medium · Data retention changed · 2026-08-25 · privacy_policy
Wispr AI removed its disclosure of "zero data retention agreements with third-party AI providers" for Flow Dictation models, eliminating explicit commitments that customer data would not be retained by third-party AI providers for model training purposes. This represents a material weakening of data retention commitments previously disclosed to customers.
high · Training rights changed · 2026-08-19 · tos
Wispr AI renamed "Privacy Mode" (which disabled training use of customer content) to "Improve the model for everyone" and inverted its meaning to enable training use of customer content by default. The definition now states this feature "enables the use of Customer Content for evaluating, training, or improving Wispr's AI models," reversing the previous opt-out model to an opt-in model for training rights.
high · Training rights changed · 2026-08-19 · privacy_policy
Wispr AI shifted from a Privacy Mode opt-in model where "zero dictation data is used for AI model training" to a new "Improve the Model for Everyone" feature that allows Wispr to use customer dictation data (audio, transcripts, edits) for model training by default, with opt-out control. This represents a material change in training rights, moving from no training by default to training enabled unless explicitly disabled.
medium · Terms changed · 2026-08-19 · trust_center
Wispr AI's trust document was substantially reorganized and rewritten for clarity, with multiple material changes to security and data-handling disclosures. Key changes include: removal of named individual security personnel (replaced with role-based descriptions), significant restructuring of Privacy Mode and Cloud Sync controls with clarified enforcement mechanics, removal of the API export capability (now sunsetted), and consolidation of HIPAA BAA revocation procedures. The document also removes specific details on local-data retention policies and simplifies enterprise control descriptions, though the underlying commitments remain substantively similar.
medium · AI feature launched · 2026-08-05 · ai_disclosure
Wispr AI announced the launch of Wispr Flow Notetaker, an AI meeting assistant that captures transcripts, identifies speakers, and generates summaries and next steps from conversations. This represents a newly disclosed AI-powered feature that processes customer content (meeting audio/conversations) to generate derived outputs.
medium · AI feature launched · 2026-08-11 · ai_disclosure
Wispr AI announced the launch of Wispr Flow Notetaker, a new AI meeting assistant feature that captures transcripts, identifies speakers, and generates summaries and next steps from conversations. This represents a newly disclosed AI-powered feature processing customer content (meeting conversations).
Paste your vendor list, confirm your email, and our engine builds your report automatically — which vendors add AI subprocessors or reserve the right to train on your data, and where your nth-party exposure concentrates. Free, delivered during your working day.