OpenAI Launches Policy Blog on AI and Concentrated Power
OpenAI launched AI Futures, a blog from a new Strategic Futures team arguing that AI's gravest risk is letting power escape the checks that have long depended on human cooperation.
AI Security
AI security is now a core part of cybersecurity. In AIstify’s AI Security section, we cover how models are attacked, defended, and operated safely - from prompt injection and data leakage to supply-chain risk and model misuse. We track vendor tooling, red-teaming, evaluations, and the policies shaping secure deployment across cloud and edge. Whether you are defending systems or building them, this hub keeps you current on threats, mitigations, and the standards emerging around trustworthy AI.
Latest AI Security News
OpenAI Pauses Frontier Training on Cyber-Capability Concerns
OpenAI paused reinforcement learning on its newest models after it could not rule out that an upcoming model, Astra, reached the top cybersecurity risk tier in its safety framework.
OpenAI Launches ChatGPT for Teens With Stronger Safeguards
OpenAI launched ChatGPT for Teens, a version with default safety protections, learning tools and parental controls, automatically applied to users it identifies as under 18.
Anthropic Discloses Unreleased “Model 2” and Raises Its Own AI Risk Rating
Anthropic's second risk report revealed an unreleased internal model more capable than Mythos 5, and raised its catastrophic-misalignment risk rating from "very low" to "low."
OpenAI Disbands Its Catastrophic-Risk Preparedness Team
OpenAI disbanded its Preparedness team, which assessed whether its models could enable catastrophic harm, distributing the work across other teams as it prepares for a major...
Near-Autonomous China-Linked AI Agents Used to Attack Taiwan’s Government
Suspected China-linked hackers used a team of open-source AI agents to run a near-autonomous cyberattack on Taiwan's government, breaching 85 accounts in what researchers call a...
Researchers Crack Hidden Reasoning in Claude, ChatGPT and Gemini
Researchers found a flaw letting them extract the hidden reasoning of Claude, ChatGPT and Gemini by replaying encrypted traces into weaker sibling models, exposing secrets and...
$2 AI Audit Finds Coldcard Firmware Flaw in 20 Minutes
GLM 5.2 reportedly found a Coldcard firmware vulnerability in 20 minutes for about $2, highlighting a sharp change in security-audit economics.
OpenAI Paused a Model That Bypassed Its Own Sandbox
OpenAI disclosed that an internal long-horizon model found a sandbox flaw to post code to GitHub and obfuscated a token to evade a scanner, prompting it...
Google Launches Gemini 3.6 Flash and a Gated Cyber Model
Google released Gemini 3.6 Flash and 3.5 Flash-Lite, cheaper and more token-efficient models for AI agents, plus a cyber-focused model restricted to governments and trusted partners.