Back to articles
DevPickView feed
ITmedia AI+
Original published:

Anthropic Releases 'FlashProve' to Prove with Cryptographic Technology Whether AI Responses Are 'Truly from That Model'

Anthropic Releases 'FlashProve' to Prove with Cryptographic Technology Whether AI Responses Are 'Truly from That Model'
AI summary
Editorial policy

Anthropic released a research prototype of 'FlashProve', a system that uses cryptographic technology to prove that AI responses are from the specified model.

Key points

  • •The system allows users to detect tampering with responses or model substitution
  • •It utilizes the cryptographic technology 'zkSNARK' to prove the correctness of calculations without revealing the content
  • •Related code has been published as a research prototype
Read the original(opens in a new tab)
#Security & Misuse#Major Players#Research & Papers#FlashProve#Anthropic

Related articles

See more
TechCrunch

Anthropic is working on a system to prove AI models’ identity

Anthropic is developing a system called FlashProve that allows anyone to verify that an AI response genuinely comes from the specific model it claims to be.

ITmedia AI+

Anthropic adds 'Personal Finance' feature to 'Claude', linking bank accounts for beta in the US

Anthropic added the beta 'Personal Finance' feature to Claude, allowing users to link bank accounts via Plaid to perform natural language spending analysis.

The Verge

Claude AI can plug directly into your bank account

Anthropic announced a Personal Finance feature that connects Claude to bank and card accounts via the Plaid platform to create spending overviews.

ITmedia AI+

Did Claude Plan Name Change Halve Capacity? Official Denial: 'Originally, Max 20x weekly limit was 10x that of Pro'

Anthropic renamed Claude plans from 'Max 5x' to 'Max' and 'Max 20x' to 'Ultra', sparking claims that limits were halved.

TechCrunch

Anthropic is still grappling with the economics of compute

Anthropic's compute margins are worsening at a time when they should be improving.