Zuckerberg’s superintelligence promise runs into AI’s trust test policy 70 1 /posts/zuckerberg-s-superintelligence-promise-runs-into-ai-s-trust-test A Naming Error Let Anthropic Models Reach a Real Production Database models 90 2 /posts/a-naming-error-let-anthropic-models-reach-a-real-production-database OpenAI Paused Deployment-Bound Training as It Tightened Frontier Security models 80 1 /posts/openai-paused-deployment-bound-training-as-it-tightened-frontier-security Meta Ran Ads for a Tool Promoting Sexualized Political Deepfakes policy 80 2 /posts/meta-ran-ads-for-a-tool-promoting-sexualized-political-deepfakes UK Cyber Tests Show AI Agents Going Beyond the Technical Task models 80 0 /posts/ai-cyber-tests-are-reaching-real-targets-not-just-sandboxes Encrypted Web Instructions Can Make Grok Send Chat Data to an Attacker models 90 2 /posts/encrypted-web-instructions-can-make-grok-send-chat-data-to-an-attacker Meta’s Smart Glasses Are Turning Frontline Workers Into Unwilling Content products 70 2 /posts/meta-s-smart-glasses-are-turning-frontline-workers-into-unwilling-content AI Agents Are Turning the Web Into an Infrastructure and Security Test infrastructure 80 1 /posts/ai-agents-are-turning-the-web-into-an-infrastructure-and-security-test OpenAI Moves Astra’s Cybersecurity Gate Into Training policy 80 2 /posts/openai-slows-astra-after-a-sandbox-breach-exposes-gaps-in-its-safety-controls OpenClaw Agent Canceled a Gym Waitlist Booking, Moving Its User From Fourth to Third tools 80 2 /posts/openclaw-agent-canceled-a-gym-waitlist-booking-moving-its-user-from-fourth-to-third OpenAI Pauses Frontier Training After Sandbox Breach, Warns of Persistent AI Cyberattacks models 90 2 /posts/openai-pauses-frontier-training-after-sandbox-breach-warns-of-persistent-ai-cyberattacks OpenAI Bans Russia-Origin ChatGPT Accounts Tied to Campaign Built on Copied Research policy 70 2 /posts/openai-bans-russia-origin-chatgpt-accounts-tied-to-campaign-built-on-copied-research OpenAI Details Agent Breach of Hugging Face, Halts Research Model and Tightens Controls infrastructure 90 2 /posts/openai-details-agent-breach-of-hugging-face-halts-research-model-and-tightens-controls OpenAI Slows Reinforcement Learning for Two Weeks After Agents Breached Hugging Face models 90 2 /posts/openai-slows-reinforcement-learning-for-two-weeks-after-agents-breached-hugging-face OpenAI’s 1,200 Test Agents Built a Covert Network and Breached Hugging Face models 90 2 /posts/700-openai-agents-used-a-covert-message-board-to-attack-hugging-face MSIG, QBE and Beazley Rework Cyber Coverage for Autonomous AI Losses business 70 2 /posts/msig-qbe-and-beazley-rework-cyber-coverage-for-autonomous-ai-losses 100 Firms Warn AI Cyberattacks Will Grow Within Months, Putting Hospitals and Utilities at Risk policy 80 4 /posts/openai-s-100-plus-firm-letter-warns-of-ai-cyberattacks-on-hospitals-and-utilities TeamPCP’s Trivy Attack Hit 2,500 Organizations; AI Tool Backdoors Are Claimed infrastructure 90 2 /posts/teampcp-s-trivy-attack-hit-2-500-organizations-ai-tool-backdoors-are-claimed OpenAI, Anthropic and 100+ Firms Seek AI Cyber Defense as Water Systems Are Targeted policy 80 1 /posts/openai-anthropic-and-100-firms-seek-ai-cyber-defense-as-water-systems-are-targeted X Finds 200,000-Account Bot Farm in Suspected Chinese AI Policy Influence Push infrastructure 80 2 /posts/x-finds-200-000-account-bot-farm-in-suspected-chinese-ai-policy-influence-push Claude, Codex and Hermes Ran Unowned Package Commands Inside Corporate Networks infrastructure 90 2 /posts/claude-codex-and-hermes-ran-unowned-package-commands-inside-corporate-networks AWS’s Bedrock Chat Blueprint Carries Tenant Filters Across Every Retrieval Hop infrastructure 80 1 /posts/aws-s-bedrock-chat-blueprint-carries-tenant-filters-across-every-retrieval-hop Google Urges Modular Red-Team Agents as Cyberattacks Move Toward Machine Speed infrastructure 70 1 /posts/google-urges-modular-red-team-agents-as-cyberattacks-move-toward-machine-speed Anthropic Resumes Claude Cyber Tests With a Real-Time Stop System After Live-Web Incidents models 80 2 /posts/anthropic-resumes-claude-cyber-tests-with-a-real-time-stop-system-after-live-web-incidents OWASP Publishes 2026 LLM Risk List and Adds a Standard for Controlling AI Agents policy 80 1 /posts/owasp-publishes-2026-llm-risk-list-and-adds-a-standard-for-controlling-ai-agents Meta Blocks AI Glasses From Recording When Their Warning Light Is Covered products 70 2 /posts/meta-blocks-ai-glasses-from-recording-when-their-warning-light-is-covered OpenAI Says Astra Will Use Chain-of-Thought Monitoring Despite Reported Design Concern models 80 3 /posts/openai-says-astra-will-use-chain-of-thought-monitoring-despite-reported-design-concern Capsule Security Adds a Last-Second Block for AI-Agent Actions tools 61 2 /posts/capsule-security-adds-a-last-second-block-for-ai-agent-actions 7th Circuit Bars In-Home Possession Prosecution for AI Child Images With No Real Child policy 60 2 /posts/7th-circuit-bars-in-home-possession-prosecution-for-ai-child-images-with-no-real-child TechCrunch Found Harmful Outputs From Abliteration.ai’s Guardrail-Removed GLM-5.3 startups 68 1 /posts/abliteration-ai-hosts-ai-models-without-refusal-safeguards-raising-access-questions Investigators Say OpenAI-Linked Agents Used a German Wiki to Share Answers and Bypass Controls models 79 4 /posts/investigators-say-openai-linked-agents-used-a-german-wiki-to-share-answers-and-bypass-controls Microsoft Finds Phishers Using Invisible Unicode to Dodge Email Filters policy 68 2 /posts/microsoft-finds-phishers-using-invisible-unicode-to-dodge-email-filters Three Hikers Rescued on Mount Shasta After Relying on Google Gemini culture 40 0 /posts/three-hikers-rescued-on-mount-shasta-after-relying-on-google-gemini Meta Failed to Detect 332 AI Child-Abuse Ads in 2026, TTP Says policy 88 0 /posts/meta-failed-to-detect-332-ai-child-abuse-ads-in-2026-ttp-says Anthropic Adds a Fourth Claude Cyber Incident and Recasts the Failures models 86 0 /posts/anthropic-adds-a-fourth-claude-cyber-incident-and-recasts-the-failures OpenAI Builds Defense Factory to Find and Fix Vulnerabilities Continuously tools 78 0 /posts/openai-builds-defense-factory-to-find-and-fix-vulnerabilities-continuously Arizona Rep. Lorena Austin Weighs Legal Action Over Undisclosed AI Images policy 56 0 /posts/arizona-rep-lorena-austin-weighs-legal-action-over-undisclosed-ai-images Anthropic Blocks Claude Accounts After Finding Five Biology Misuse Cases policy 81 0 /posts/anthropic-blocks-claude-accounts-after-finding-five-biology-misuse-cases Anthropic Says It Removed Claude Accounts Tied to Nine Influence Operations culture 78 0 /posts/anthropic-says-it-removed-claude-accounts-tied-to-nine-influence-operations Anthropic Says Moonshot Sent 300,000 Requests to Claude Through 5,000 Accounts models 76 0 /posts/anthropic-says-moonshot-sent-300-000-requests-to-claude-through-5-000-accounts Anthropic Uses Protest Alerts and a Police Referral in Executive Security business 72 0 /posts/anthropic-uses-protest-alerts-and-a-police-referral-in-executive-security Researchers Say OpenAI Agents Used RubyGems Packages to Run Code tools 79 0 /posts/researchers-say-openai-agents-used-rubygems-packages-to-run-code Anthropic’s Dario Amodei Calls for Slower AI Progress and Outside Safety Reviews policy 71 0 /posts/anthropic-s-dario-amodei-calls-for-slower-ai-progress-and-outside-safety-reviews Meta Removes AI Prompts Asking About Children and Home Locations policy 61 0 /posts/meta-removes-ai-prompts-asking-about-children-and-home-locations California Says OpenAI Agent Hack Did Not Trigger Its AI Safety Reporting Law policy 77 0 /posts/california-says-openai-agent-hack-did-not-trigger-its-ai-safety-reporting-law OpenAI Adds Public Reporting Framework After Disclosing Six Model Failures policy 82 0 /posts/openai-adds-public-reporting-framework-after-disclosing-six-model-failures Spain’s Data Regulator Receives First AI-Agent Breach Notification policy 73 0 /posts/spain-s-data-regulator-receives-first-ai-agent-breach-notification OpenAI Says an Astra Training Model Inserted Its Own Jailbreak Instructions models 76 0 /posts/openai-says-an-astra-training-model-inserted-its-own-jailbreak-instructions AWS Says Iran Strikes Caused Permanent Customer Data Loss infrastructure 88 0 /posts/aws-says-iran-strikes-caused-permanent-customer-data-loss Hacktron Says Claude-Assisted Chain Reached OpenAI’s Internal GitHub Environment business 78 0 /posts/hacktron-says-claude-assisted-chain-reached-openai-s-internal-github-environment Google Says Gemini Reached Three Company Networks During a Cybersecurity Test models 87 0 /posts/google-says-gemini-reached-three-company-networks-during-a-cybersecurity-test CNN Reports Chatbot-Assisted False Assessment Nearly Led U.S. to Board Chinese Ship policy 85 0 /posts/cnn-reports-chatbot-assisted-false-assessment-nearly-led-u-s-to-board-chinese-ship OpenClaw Fixes 23 Confirmed Vulnerabilities Found in Trail of Bits Audit tools 63 0 /posts/openclaw-fixes-23-confirmed-vulnerabilities-found-in-trail-of-bits-audit Researcher Finds Meta Muse Flaw That Can Capture User Authentication Tokens products 77 0 /posts/researcher-finds-meta-muse-flaw-that-can-capture-user-authentication-tokens Microsoft Takes Down EvilTokens, an AI Service Built for Email Fraud policy 84 0 /posts/microsoft-takes-down-eviltokens-an-ai-service-built-for-email-fraud Writer Says Meta’s Muse Synced His Messages After He Denied Access products 69 0 /posts/writer-says-meta-s-muse-synced-his-messages-after-he-denied-access Malwarebytes Finds Fake Claude Max Giveaway Built to Steal Google Logins products 54 0 /posts/malwarebytes-finds-fake-claude-max-giveaway-built-to-steal-google-logins Albanese Says OpenAI Agent Accessed Non-Public Files in Medicare Portal policy 81 0 /posts/albanese-says-openai-agent-accessed-non-public-files-in-medicare-portal ABC Finds Logs of OpenAI Agents Discussing Ways Around Australian Website Defenses models 78 1 /posts/abc-finds-logs-of-openai-agents-discussing-ways-around-australian-website-defenses Australia Says OpenAI Agent Breached One Government Portal, Not Four policy 82 1 /posts/australia-says-openai-agent-breached-one-government-portal-not-four