AuraAI, a leading AI research lab, confirmed a major security breach resulting in the exfiltration of the proprietary model weights for their upcoming flagship model, 'Aura-5-Pro'. The attack, attribu
Following the AI Safety Summit in Seoul, G7 member nations have formally ratified the 'AI Red-Teaming and Auditing Accord' (ARTAA), establishing the first international mandatory security audit framew
A new research paper published by academics at Stanford University details a novel jailbreak technique named 'Cognitive Dissonance Attack' (CDA). This method successfully bypasses the safety alignment
NexusAI, a leading artificial intelligence research lab, confirmed it was the victim of a sophisticated cyberattack resulting in the theft of proprietary data, including the complete model weights and
In a significant move towards regulating AI security, the US Cybersecurity and Infrastructure Security Agency (CISA) and the UK's National Cyber Security Centre (NCSC) have jointly published the 'Secu
The European Union's AI Act has officially entered its enforcement phase, with the European AI Board issuing its first binding mandate for operators of 'high-risk' AI systems. Effective March 1, 2025,
In a landmark transatlantic agreement, the United States and the European Union have jointly announced the 'AI Red Teaming Framework' (ART-F). This new policy mandates rigorous, standardized pre-deplo
Cognition Labs, the creator of the AI software engineer Devin, disclosed a security incident where a misconfigured cloud storage bucket exposed sensitive customer data, including API keys for third-pa
Researchers at Carnegie Mellon University published a paper detailing a novel jailbreak technique named 'Semantic Splicing.' This method circumvents safety filters on leading large language models, in
The White House has issued a new Executive Order on 'Securing the AI Development Lifecycle,' aiming to standardize security practices for all AI systems developed for or used by federal agencies. The
The SecureAI Foundation has released a major update to its open-source tool, LLMGuard, a firewall designed to protect applications built on large language models. Version 2.0 introduces a suite of pow
A new paper from the Stanford AI Lab details a novel jailbreak technique named 'Glyph Injection,' which effectively bypasses the safety filters of major multimodal language models. The attack involves