AI safety & Anthropic

Automated Subtle Emergent Misalignment Research (opens in a new tab)

Automated Subtle Emergent Misalignment Research Anthropic Alignment Science blog

Kosa8 Labs Publishes AI Agent Security Benchmark for Regulated Firms (opens in a new tab)

Hackers hijacked a Claude Code extension, exposing yet another new AI security risk (opens in a new tab)

Anthropic Says Claude Built Five Exploit Chains Across Major Software (opens in a new tab)

'It’s 100 percent a national security threat': Accenture contractor connected an AI agent to the Pentagon’s secure network, and then sent data to a high-risk website (opens in a new tab)

Agentic AI Security: Why 90% of Enterprises Face Risks (opens in a new tab)

Anthropic staff warn AI could kill us all as its safety rules weaken (opens in a new tab)

SecurityHQ Research Reveals AI Agent Accountability Gap Across Enterprise Security Operations (opens in a new tab)

Anthropic Security Breach Exposes Internal Systems in Research Test (opens in a new tab)

Cursor AI Says Co Identified Security Incident Affecting A Modified Claude Code Versus Code Extension Impacting At Most 5,000 Users (opens in a new tab)

Trust3 AI brings enterprise agent security to Claude Code (opens in a new tab)

Datadog: Agentic AI will reshape observability and security in Southeast Asia (opens in a new tab)

Pentagon AI Security Breach: Accenture Employee Connected Personal Claude Agent, Sensitive Data Flowed Out for Two Months Before Detection (opens in a new tab)

IBM (IBM) Debuts AI Agent Security Controls For Hybrid Cloud (opens in a new tab)

Anthropic-Themed Phishing Emails Bypass Top Security Tools (opens in a new tab)

Mintive Tops 2.5 Million Daily Agent Calls as Enterprises Make AI Agent Identity a Security Priority (opens in a new tab)

Opinion | Anthropic’s IPO Will Be a National-Security Win for Taiwan (opens in a new tab)

Claude AI Exploits Linux Kernel Flaw to Execute Root Commands Without User Interaction (opens in a new tab)

Anthropic AI Biology Lab Opens as Safety Concerns Grow (opens in a new tab)

Natural Emergent Misalignment from Reward Hacking in Production RL [pdf] (opens in a new tab)

Sen. Warren probes AI safety, market risks in Anthropic IPO (opens in a new tab)

Claude — Elevated errors affecting the Google Calendar connector (opens in a new tab)

En cours · We are investigating elevated errors affecting the Google Calendar connector. Some requests to list or search calendar events have been failing since October 26. We will provide an update as soon as possible.

Enterprises balance AI agent autonomy with security guardrails (opens in a new tab)

Autonomous AI agent discovers critical vulnerabilities in EPICS software (opens in a new tab)

AI agent safety will be the 'issue of our time': Cisco president (opens in a new tab)

Claude AI Builds One-Click Chrome and 0-Click Linux Exploit Chains, Bypasses Android Lock Screen (opens in a new tab)

Black Duck launches AI agent to fix vulnerabilities as fast as attackers exploit them (opens in a new tab)

Claude is the best security software you aren't using (opens in a new tab)

Why Agentic AI Belongs in Your Employee Security Framework (opens in a new tab)

Claude — Elevated errors affecting the Gmail connector (opens in a new tab)

En cours · We are investigating elevated errors when searching and reading email with the Gmail connector. This issue does not affect other connectors. We will provide an update as soon as possible.

Agentic AI linked to rise in cyberattacks, Canada’s cybersecurity agency says (opens in a new tab)

What CIOs can learn from Anthropic’s safety pullback (opens in a new tab)

Discovering and exploiting vulnerabilities in real-world software with Claude (opens in a new tab)

Lumia Security Adds Claude Compliance API Integration, Expanding AI Governance Options for Enterprises (opens in a new tab)

Discovering and exploiting vulnerabilities in critical, real-world software with Claude (opens in a new tab)

Dell ties supply chain security to agentic AI risks (opens in a new tab)

Claude — Elevated errors for Claude Sonnet 5.5 (opens in a new tab)

Résolu · The issue affecting claude.ai, the Claude API, Claude Code, and Claude Cowork has been resolved. Impact occurred from 07:42 PT / 14:42 UTC to 08:17 PT / 15:17 UTC.

How AI Agent Security is creating a Password Apocalypse (opens in a new tab)

SurgeONE.ai Unveils Avery Actions Agentic AI for Compliance, Cybersecurity and Data Operations (opens in a new tab)

Claude — Elevated errors for Claude Sonnet 5.5 (opens in a new tab)

Résolu · We are investigating elevated errors on requests to Claude Sonnet 5.5. We will provide an update as soon as possible.

Anthropic’s Latest Claude Security Disclosure Will Make You Sympathize With Claude (opens in a new tab)

Cloudflare's Agentic AI Strategy: Beyond Cybersecurity (opens in a new tab)

AGNTCon + MCPCon 2026: Agentic AI Foundation Adds Four Projects as MCP Turns to Security (opens in a new tab)

IPO: Anthropic set to make stock market debut as questions swirl about safety of AI (opens in a new tab)

CodeRabbit Commits $1M in Free Security Scans (opens in a new tab)

Scoop: The FTC is demanding Anthropic execs testify in expanding AI safety probe (opens in a new tab)

Microsoft Expands AI Security Platform to Address Agentic AI Security Risks (opens in a new tab)

IPO: Anthropic set to make stock market debut as questions swirl about safety of AI (opens in a new tab)