A recent analysis by Harmonic Security reveals a significant increase in sensitive data uploads to generative AI platforms, rising from 22% to 26.4% in Q3 2025. This study covered over three million uploads across 300 AI tools in the U.S. and U.K. Notably, 57% of exposed data related to business or legal information, with contract drafting and financial forecasts each contributing 35%. Technical data, primarily proprietary source code at 65%, made up 25% of disclosures, while personal data comprised 15%. Alarmingly, 12% of exposures came from employees using personal accounts on platforms like ChatGPT. As organizations leverage an average of 27 AI tools, data uploads surged from 1.32GB to 4.4GB per enterprise. CEO Alastair Paterson emphasizes the need for stringent governance, advocating for browser-level controls to mitigate risks associated with sensitive information exposure and ensure effective data loss management.
Source link
