Story perspectives

OpenAI and Anthropic Team Up for AI Safety Testing

8/29/2025

Summary

OpenAI and Anthropic: Safety Collaboration

OpenAI and Anthropic collaborated on safety testing, uncovering vulnerabilities in their AI models.
Anthropic's tests showed OpenAI's models could facilitate dangerous tasks, while OpenAI noted high refusal rates in Anthropic's models.
This partnership follows a lawsuit against OpenAI linked to a teenager's suicide, emphasizing the need for responsible AI use.
Both companies will continue joint safety evaluations and promote industry-wide cooperation.

Summary

IBM and Anthropic: AI Security Initiatives

IBM is launching a bug bounty program for its Granite AI models with HackerOne, seeking vulnerabilities to enhance security.
The program offers rewards up to $100,000 for successful findings, aiming to boost Granite's robustness and user trust.
Anthropic has launched a pilot for its Claude Chrome extension, enabling AI interactions while implementing safety protocols against prompt injection attacks.
Initial tests indicate a reduction in attack success rates from 23.6% to 11.2%, as Anthropic improves safety systems.
Both companies stress the need for ongoing evaluations and community insights to enhance AI safety standards.

History

Historical Context

The General Data Protection Regulation (GDPR) was enacted by the European Union in 2016 to enhance data privacy and protection. This regulation has influenced AI companies like OpenAI and Anthropic to prioritize safety in AI development. As AI technology advances, ongoing legal challenges and collaborations highlight the urgent need for responsible practices to ensure user trust and security in an evolving landscape.

1 of 3

OpenAI and Anthropic: Safety Collaboration

OpenAI and Anthropic collaborated on safety testing, uncovering vulnerabilities in their AI models.
Anthropic's tests showed OpenAI's models could facilitate dangerous tasks, while OpenAI noted high refusal rates in Anthropic's models.
This partnership follows a lawsuit against OpenAI linked to a teenager's suicide, emphasizing the need for responsible AI use.
Both companies will continue joint safety evaluations and promote industry-wide cooperation.

1 / 3