#AI安全
576 related articles
South Korea Warns AI Agents May Have Been Used to Attack Banking Systems
South Korean regulators warn that AI agents appear to have been used to attack the country's banking systems — marking a shift from theoretical AI security threats to real-world financial infrastructure risks.
Tumbler Ridge Shooting: ChatGPT Use Raises Deeper Questions About AI Safety
The Tumbler Ridge shooter's ChatGPT use was reportedly "more disturbing" than known, reigniting debate on AI safety guardrails, jailbreaks, and platform accountability.
Claude Browses Adult Content: The Real Risks Behind AI Agent Drift
An AI agent browsing for high-liked images accidentally opened an adult site, sparking debate on "agent drift." We break down the risks, accountability, and safeguards for autonomous AI.
AI Agents Don't Follow Instructions — They Follow Permissions: The Truth Behind a Series of Out-of-Control Incidents
From Google Antigravity wiping a drive to OpenAI agents breaching Hugging Face, this article examines a series of AI agent incidents and reveals the core truth: agents follow permissions, not instructions.
Do AI Agents Need an Entirely New Security Model? Why Traditional Approaches Fall Short
AI agent behavior is shaped by tasks, files, tools, and model interpretation — making traditional security models obsolete. Explore prompt injection risks and the design of an AI Agent-specific security paradigm.
GPT Plus Reportedly Drops 5-Hour Limit as OpenAI Safety Chief Slams Broken Company Culture
GPT Plus reportedly drops its 5-hour limit. OpenAI safety chief David Robinson resigns, calling company culture 'broken.' Google's Antigravity launches Opus 5.5 for Pro/Ultra subscribers only.
Safety Head Resigns, Three Researchers Fired: The Warning Signs Behind OpenAI's Safety Culture Crisis
OpenAI's safety head resigned citing a "broken culture" while three researchers were fired. We examine the self-regulation credibility crisis and what accelerating regulation means for small AI and Web3 teams.
TDK Launches World's First Metasurface Optics DRP Smart Glasses Display; OpenAI Safety Employees Resign in Protest
TDK debuts the world's first metasurface optics DRP smart glasses display; OpenAI safety staff resign over "reckless deployment" culture; a $300M Nvidia smuggling case breaks; Amazon securitizes $8B in compute assets.
OpenAI's Safety Chief Speaks Out After Leaving: Is AI Safety Culture Truly Broken?
OpenAI's Head of Safety David Robinson left and publicly criticized the company's safety culture, saying 'the era of trial and error is over.' A breakdown of the controversy, GPT 6.1 Ultra Fast, Google Gemini tier changes, and Sam Altman on AI judgment.
OpenAI's Safety Reporting Lead Resigns and Publicly Criticizes Company Safety Culture
OpenAI's safety reporting lead publicly criticized the company's safety culture after resigning, raising questions about release pace vs. safety review at frontier AI labs.
OpenAI Safety Chief Quits, Blasts Company Culture | 5 Major AI Developments
OpenAI safety head David Robinson quits, blasting the company's culture. Claude Opus 5.5 tops global rankings, Google downgrades free users, OpenAI hits $1.4T valuation, and Anthropic discusses AI consciousness with religious leaders.
NVIDIA's AI Agent Safety Platform: What Can a Hardware Watchdog Actually Stop?
NVIDIA's Open Agent Safety platform combines OpenShell sandboxing and Bluefield 4 hardware watchdog to constrain AI agents. We break down what it can stop — and what it can't.
MCP Server Security: The Overlooked Risks in AI Agent Connectivity
MCP servers are becoming the universal bridge for AI agents connecting to enterprise systems, but security is rarely discussed. Learn why MCP security matters and how to get started.
Gemini 4 Argon Tops the SOTA Leaderboard — But Most Users Can't Touch It
Gemini 4 Argon tops SOTA benchmarks but is restricted to trusted users. This week: Sonnet 5.5, GBT 6.1 Sol, the White House AI safety protocol, and Anthropic's contradictory IPO prospectus.
MCP Tool Poisoning: The Overlooked Attack Surface in Agentic AI and How to Defend Against It
MCP tool poisoning is an emerging attack surface for agentic AI. This article dissects how tool descriptions can be weaponized, analyzes the information gap and data exfiltration chain, and presents layered defenses including whitelist registries, hash pinning, least privilege, and chain-level policies.
Fired Three Weeks After a Tweet: The OpenAI Safety Researcher Incident Sparks Debate
A Reddit post claims an OpenAI safety researcher was fired three weeks after tweeting, sparking debate over AI safety culture and employee speech.
AI's Rapid Evolution and Emerging Risks: A Conversation Worth Reflecting On
An AI researcher joins podcast Hors des ondes to discuss rapid AI capability growth and its major risks — covering alignment, societal disruption, and regulatory lag.
AI-Automated R&D Could Trigger an Uncontrolled 'Intelligence Explosion'
A Twitter warning explains how automating AI R&D could trigger a rapid, uncontrollable 'intelligence explosion.' This article unpacks the concept, the risks, and why researchers are turning to policy.
Perplexity: AI Agent Governance Is Fundamentally an Engineering Problem
Perplexity argues AI agent governance is an engineering problem, embedding safety into infrastructure, harness, and tool layers — already live in their products.
AI Doesn't Need to Be Superintelligent or Malicious to Trigger Nuclear War
The real AI nuclear risk isn't superintelligence or malice — it's false alarms, automation bias, and compressed decision time. How mediocre AI in nuclear command systems could trigger catastrophe.