CalypsoAI Newsletter

Welcome to our newsletter! We're excited to share interesting news from the world of AI, which we trust you'll enjoy reading. Stay tuned for more "hot-off-the-press" articles and updates.

Our AISEC podcast is live!

Neil Serebryany, CalypsoAI CEO, and renowned cybersecurity expert and six-time CISO Jim Routh explore AI security, where governance, ethics, and cybersecurity converge, shaping the future of our increasingly AI-driven world.

Listen to our first episode: Decoding the Future: AI Governance and Cybersecurity First Principles, featuring our special guest Rick Howard, CSO, Chief Analyst, and Senior Fellow at the CyberWire.

How Do Companies Use Artificial Intelligence?

It seems that most of the companies dominating the headlines in the AI ecosystem—rightfully—are those that can trigger the greatest economic damage when their AI systems fail or are used against them by threat actors. But many industries have deployed AI-driven technologies successfully and profitably while also managing risk. This article explores the novel uses for AI that have been undertaken by the farming and agriculture, logistics, retail, healthcare, and automotive industries.?

China says generative AI rules to apply only to product for the public

The Cyberspace Administration of China (CAC) has released its first “interim” rules addressing generative artificial intelligence (GenAI) tools, which will go into effect on August 15. Chinese tech giants have been focusing their efforts on building tools available to enterprise and research, while the government observed the development and deployment of Western products, such as ChatGPT and others, while also banning their use. The new rules, which will apply only to products available to the general public, will include requirements for providers, such as security assessments, licensing, and content moderation and monitoring.

The Dark Side of AI: How Prompt Hacking Can Sabotage Your AI Systems

Prompt hacking is known by a variety of terms—jailbreaks, prompt injection, grandma jailbreaks—to name a few, but the goal is typically to manipulate the model in such a way that system rules are circumvented and the model provides information it should not provide, such as personal data or illegal, unethical, or dangerous instructions. This article very effectively summarizes the nuances of prompt hacking and offers defensive mechanisms to combat it.?