Prompt injection is an attack in which untrusted text or content attempts to override an AI application’s instructions. The malicious instruction may come directly from a user or indirectly from a webpage, document, email, or tool result processed by the model. If successful, it can cause data disclosure, unauthorized tool use, altered output, or bypassed policies. Because language models do not inherently distinguish trusted commands from quoted content, filtering suspicious phrases alone is insufficient. Defenses include isolating data from instructions, minimizing permissions, validating tool arguments, requiring approval for sensitive actions, restricting secrets, and testing with adversarial content. Applications should assume that any external content may contain instructions designed to manipulate the model.
A federal judge ruled that Elon Musk’s lawsuit accusing OpenAI of abandoning its charitable mission can proceed to trial, rejecting dismissal efforts by OpenAI and Microsoft.
Anthropic has partnered with Allianz to deploy its Claude AI models and custom agents across the insurer, emphasizing responsible, transparent AI for multi-step workflows and regulatory compliance.
OpenAI introduced ChatGPT Health, a dedicated experience that securely integrates personal health data with AI, helping users navigate wellness, lab results, and appointments with added privacy protections.
Lenovo introduced Qira, a system-level AI spanning PCs, tablets, smartphones, and wearables. The assistant offers context-aware, cross-device intelligence for writing, collaboration, and workflow management.
xAI closed a $20 billion Series E round, exceeding its original target and adding strategic backing from NVIDIA and Cisco. The funding supports expanded data center capacity, large-scale GPU deployment, and continued development of the Grok model family.
OpenAI is paying employees an average of $1.5 million in stock-based compensation, far exceeding peer startups as it races to retain top AI talent. The strategy is inflating losses and reshaping expectations ahead of a potential IPO.
OpenAI is hiring a Head of Preparedness to study emerging risks tied to rapidly advancing AI models, including mental health impacts and cybersecurity threats. The move reflects rising concern over how frontier capabilities could be misused as models grow more powerful.
Chinese generative AI startups MiniMax and Zhipu AI have disclosed their financials ahead of potential Hong Kong listings, highlighting modest revenues and mounting losses compared with U.S. peers.
Visa said it successfully completed hundreds of AI-driven transactions in a pilot program testing agent-based payments. The effort reflects a broader push across fintech to let AI agents handle purchases on behalf of consumers.
OpenAI released a new version of ChatGPT Images powered by a flagship image generation model that delivers faster output and more precise edits. The update also introduces a dedicated Images workspace and a lower-cost API offering, GPT Image 1.5.