When Microsoft reviewed the early results of Anthropic’s new AI model for cybersecurity, it was a sobering scene: Hundreds of bugs were identified, many of them critical.

What followed was a “mad dash” to patch them — but many more remain.

https://www.propublica.org/article/anthropic-mythos-microsoft-software-vulnerabilities?utm_source=mastodon&utm_medium=social&utm_campaign=mastodon-post

#News #Cybersecurity #Tech #Microsoft #Technology #Anthropic #AI #ArtificialIntelligence #Hacking

Writeup of the openai attack on HuggingFace via @campuscodi
Key points
- access to source: OSS code, trivially decompiled JAR files,... allows for the LLMs to perform offline searches for vulnerabilities at scale. Then, when the goal "solve this problem" could only be met by attacking an external company, it did.

Interesting hypothesis that part of the attack may have involved supplying malicious artifacts to other systems to get them to run your exploit. This is why package managers must require signed artifacts & build tools must check them

#cybersecurity

https://www.hacktron.ai/blog/here-is-how-openai-model-hacked-huggingface

Writeup of the openai attack on HuggingFace via @campuscodi
Key points
- access to source: OSS code, trivially decompiled JAR files,... allows for the LLMs to perform offline searches for vulnerabilities at scale. Then, when the goal "solve this problem" could only be met by attacking an external company, it did.

Interesting hypothesis that part of the attack may have involved supplying malicious artifacts to other systems to get them to run your exploit. This is why package managers must require signed artifacts & build tools must check them

#cybersecurity

https://www.hacktron.ai/blog/here-is-how-openai-model-hacked-huggingface