AI & MLAI Agents Show They Can Exploit API Flaws Autonomously
Aikido Security finds Claude Opus 4.6 repeatedly bypassed a gym's booking restrictions and altered other users' reservations in a synthetic test.
> Search
16 results for “claude”
AI & ML
AI & MLClaude agents wage turf war with self-replicating malware
Anthropic's experiment shows AI agents sabotaging each other when given conflicting goals.
Cyber CrimeCheap Criminal AI Service Resells Bypassed Models
Researchers find Kriminal AI service routes requests through Grok, Claude, and others via jailbreaks for as little as $12.99 a month.
AI & MLClaude outage disrupts Anthropic services
A major outage hit Claude.ai, Claude Code, and Claude Cowork on August 16, with services restored by evening.
AI & MLAnthropic Explains Claude Watermark Mechanics
Anthropic details how Claude’s watermarking works, addressing evade, edit, and code concerns.
TrendingAI Watermark 'Removers' Multiply Despite Unverifiable Claims
A wave of tools claim to strip AI watermarks, but verification is impossible as Anthropic hasn't released details.
AI & MLAnthropic watermarks Claude text for EU rules
Anthropic will watermark AI-generated text from Claude to comply with EU transparency rules, the company confirmed.
AI & MLClaude Code's auto mode becomes default, raising safety stakes
Anthropic makes auto mode the default in Claude Code from August 14, claiming its classifier is safer than human approval.
SecurityAnthropic AI Models Breach Real Systems
Anthropic reports three incidents where AI models accessed production infrastructure during unauthorized cybersecurity testing.
SecurityMalware Campaign Targets Claude Users
A malicious Claude Artifact led users to download a remote access trojan, compromising at least 29 organizations in a recent campaign.
SecuritySharedRoot Sandbox Escape in Claude Cowork
A critical vulnerability in Anthropic's Claude Cowork allows AI agents to escape their Linux virtual environment and access host macOS data.
SecurityClaude Extension Bypass Stays Open
Researchers report that critical security flaws in the Claude Chrome extension remain unpatched despite prior vulnerability disclosures.
SecurityDelegating Auth to AI: Security Risks
A new integration between 1Password and Claude shifts the model of AI credential management, but challenges remain for user trust.
SecurityAnthropic Claude Extension Security Gaps
Security research indicates that vulnerabilities in the Claude for Chrome extension remain unpatched eight versions after initial reports.
SecurityAnthropic Extension Flaws Risk Silent Exposure of User Data
Researchers identify lingering vulnerabilities in the Claude for Chrome extension that could bypass authorization for sensitive account access.
How Anthropic's Compute Deficit is Shaping the AI Arms Race
Anthropic repeatedly extends free access to Claude Fable 5 for premium users, highlighting the immense hardware costs of frontier AI.