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.
OpenAI CEO Sam Altman said the creator of the viral AI agent OpenClaw is joining the company, while the project will continue as an open source initiative supported by OpenAI.
The Pentagon is pressing leading AI companies to make their tools available on classified military networks, seeking fewer usage restrictions as it expands AI deployment across defense operations.
OpenAI introduced Frontier, a platform enabling enterprises to build, deploy, and manage AI agents across business workflows. The system turns AI into practical coworkers that learn, act, and improve over time.
ElevenLabs CEO Mati Staniszewski says voice is emerging as the primary way people will interact with AI, as models combine speech with reasoning and agentic capabilities.
A new platform called Rent a Human allows AI agents to outsource tasks to real people when automation falls short, highlighting an unusual hybrid model of human-in-the-loop labor.
Artificial intelligence has become the top investment theme for global family offices, while cryptocurrencies remain largely sidelined, according to JPMorgan’s latest global survey.
SpaceX has acquired Elon Musk’s AI company xAI, combining rockets, satellites, and artificial intelligence into a vertically integrated effort aimed at scaling AI compute beyond Earth.
Moltbook, built on OpenClaw agentic AI, lets bots interact and form communities. Experts warn of security risks and governance challenges with AI-driven social networks.
A coalition of nonprofits urges federal suspension of xAI’s Grok AI, highlighting nonconsensual image generation, bias, and potential national security threats.
Google DeepMind has introduced AlphaGenome, a new AI model designed to analyze long DNA sequences and predict how genetic variations may influence disease development.