Discover quality AI tools
Bring together the world’s leading artificial intelligence tools to help improve work efficiency
Articles published by WilliamCarter
OpenAI has launched a preview version of Private Safety Processing for select customers, an upgraded safety solution built on its existing zero data retention policy. It uses automated systems to monitor potential misuse of AI models across multiple conversations without storing customer data, enabling detection of malicious activities such as cyberattacks and code generation even when requests are split across sessions. The system generates alerts automatically, allowing OpenAI to assess risks and contact customers if further action is needed (with customers able to decide whether to share additional data). This service addresses the challenge of balancing privacy protections with security requirements for enterprises that refuse to share sensitive information. It also responds to Anthropic’s recent policy allowing 30 days of user data retention for certain models. As AI companies compete in capabilities, security, and commercialization, privacy and security governance have become key differentiators, with both firms preparing for IPOs.
OpenAI has launched a preview version of Private Safety Processing for select customers, an upgraded safety solution built on its existing zero data retention policy. It uses automated systems to monitor potential misuse of AI models across multiple conversations without storing customer data, enabling detection of malicious activities such as cyberattacks and code generation even when requests are split across sessions. The system generates alerts automatically, allowing OpenAI to assess risks and contact customers if further action is needed (with customers able to decide whether to share additional data). This service addresses the challenge of balancing privacy protections with security requirements for enterprises that refuse to share sensitive information. It also responds to Anthropic’s recent policy allowing 30 days of user data retention for certain models. As AI companies compete in capabilities, security, and commercialization, privacy and security governance have become key differentiators, with both firms preparing for IPOs.
Cisco has open-sourced Antares, a small AI model family for code repository security scanning. It can locate known vulnerabilities, scanning 500 repositories in 15 minutes at under $1. Running locally protects sensitive source code. The 350M and 1B versions are available for verified defenders; the 3B version will be integrated into Cisco's products.
Cisco has open-sourced Antares, a small AI model family for code repository security scanning. It can locate known vulnerabilities, scanning 500 repositories in 15 minutes at under $1. Running locally protects sensitive source code. The 350M and 1B versions are available for verified defenders; the 3B version will be integrated into Cisco's products.
Google Chrome is downloading a 4GB local AI model file for Gemini Nano features without user consent, reducing disk space and raising privacy concerns. The files re-download if deleted unless a specific experimental flag is disabled. A privacy advocate has accused Google of violating EU regulations by pushing the files without explicit user knowledge. Google has not yet responded in detail.
Google Chrome is downloading a 4GB local AI model file for Gemini Nano features without user consent, reducing disk space and raising privacy concerns. The files re-download if deleted unless a specific experimental flag is disabled. A privacy advocate has accused Google of violating EU regulations by pushing the files without explicit user knowledge. Google has not yet responded in detail.
