OpenAI seeks to one-up Anthropic with new customer privacy protections
OpenAI has announced a preview of a new privacy tool called "Private Safety Processing" to monitor for misuse without saving enterprise customer data. The feature aims to counter Anthropic's policy from July, which lets the rival retain user data for 30 days on "covered models" like its Mythos-class models. This move comes as competition intensifies between the two firms, with Anthropic reportedly reaching an annualized revenue run rate of $65 billion.
Key Takeaways
- OpenAI is testing a feature named "Private Safety Processing" with select customers to detect bad actors across multiple conversations.
Unlike standard Zero Data Retention setups that scan single sessions, this method uses automated agents to spot multi-session abuse, such as malware creation attempts, without human review of conversations.
- When triggered, the system alerts OpenAI with a "narrowly defined signal", leaving data sharing to the customer's discretion.
This strategy directly contrasts with Anthropic's data-retention policy introduced in July.
- While designed to catch abuse, the policy has raised concerns among enterprises handling sensitive information.
The rival offerings underscore a growing commercial battle between the two artificial intelligence leaders as both eye initial public offerings.
- Building effective long-horizon monitoring without violating user privacy remains a critical design challenge for developers building secure AI infrastructure.
OpenAI is previewing "Private Safety Processing", an automated system that monitors multi-session misuse without storing customer data.
- Anthropic's annualized revenue run rate has reportedly reached $65 billion as it prepares for a potential IPO.
Stats & Key Facts
- #The feature aims to counter Anthropic's policy from July, which lets the rival retain user data for 30 days on "covered models" like its Mythos-class models.
- #This move comes as competition intensifies between the two firms, with Anthropic reportedly reaching an annualized revenue run rate of $65 billion.
- #Anthropic keeps session data for 30 days on "covered models", such as its Mythos-class models and Fable, allowing human inspection via a "controlled access path" logged in a "tamper-proof log".
- #OpenAI experienced slower growth in Q2 compared to Anthropic, whose annualized revenue run rate reached $65 billion and could support a valuation of $2 trillion upon listing.
OpenAI is testing a feature named "Private Safety Processing" with select customers to detect bad actors across multiple conversations. Unlike standard Zero Data Retention setups that scan single sessions, this method uses automated agents to spot multi-session abuse, such as malware creation attempts, without human review of conversations. When triggered, the system alerts OpenAI with a "narrowly defined signal", leaving data sharing to the customer's discretion.
This strategy directly contrasts with Anthropic's data-retention policy introduced in July. Anthropic keeps session data for 30 days on "covered models", such as its Mythos-class models and Fable, allowing human inspection via a "controlled access path" logged in a "tamper-proof log". While designed to catch abuse, the policy has raised concerns among enterprises handling sensitive information.
The rival offerings underscore a growing commercial battle between the two artificial intelligence leaders as both eye initial public offerings. OpenAI experienced slower growth in Q2 compared to Anthropic, whose annualized revenue run rate reached $65 billion and could support a valuation of $2 trillion upon listing. Building effective long-horizon monitoring without violating user privacy remains a critical design challenge for developers building secure AI infrastructure.
For more details please read the original article at TechCrunch AI.
Continue Learning
Comments
Comments appear only after moderation. Your email identifies your submission to the moderator and is never displayed here.
No approved comments yet.