Poland’s data watchdog UODO addresses concerns over user data collection in the ‘Feed Bielik’ AI training campaign by InPost.
UODO Responds to Data Concerns
Polish AI model Bielik AI, created by SpeakLeash Foundation, has been available in the InPost Mobile app since December. Users train the model via chat through the ‘Feed Bielik’ campaign. DGP previously noted the lack of a privacy policy and separate terms for this campaign, prompting a query to UODO.
No Administrative Proceedings Yet
UODO stated it cannot take a definitive stance without an administrative decision, as neither Bielik AI nor the ‘Feed Bielik’ service is under investigation. Regardless, data processing must comply with RODO principles, including legal bases like consent or contractual necessity.
Cloud Act and Data Transfers
User prompts and responses are stored in InPost’s private EU cloud, with frontend services using Google Cloud’s European data centers. Google clarified it lacks direct access to data, but lawyer Paweł Litwiński highlighted risks under the U.S. Cloud Act, which may compel data disclosure to American authorities.
Privacy Policy Gaps
Lawyer Tomasz Zalewski pointed out InPost’s privacy policy lacks specific mention of ‘Feed Bielik,’ though it broadly covers AI chat services. UODO acknowledged the policy includes information obligations but stressed a full assessment requires administrative verification.
User Rights Under RODO
Individuals may file complaints with UODO if they believe data is mishandled or seek compensation for damages. UODO also emphasized mandatory privacy assessments, including DPIA, for high-risk AI systems to address potential data exposure risks.
