Every event found in the daily scan — drivers that passed the filter and additional items. Filter by source category.
2026-05-02Frontier AI Labscapability
+0.0Anthropic — Claude Opus 4.7 GA + new Constitution
Opus 4.7 ל-GA — מוביל ב-Arena Elo (1503), קלט תמונות עד 2576px (פי 3+), חוקה חדשה לסירוב משימות זדוניות.
2026-05-02Incident Databasesbypass
+0.2Ynet — AI-powered investment scam in Israel using deepfakes of public figures
קמפיין דיפ-פייקים בישראל עם נגיד בנק ישראל, ראש הממשלה, גדות וגולן — מפנה לקבוצות WhatsApp של רמאים.
2026-05-02Incident Databasesbypass
+0.0Ynet — Deepfake video scams emerge as a major cyber threat in Israel
סקירה רחבה של אקוסיסטם דיפ-פייק בישראל — מקרה אחד מתוך שורה של דיווחים.
2026-05-02Policy & Standardsbypass
+0.0EU AI Act — August 2, 2026 enforcement countdown (only 8 of 27 states ready)
פחות מ-3 חודשים לאכיפה מלאה של EU AI Act על מערכות High-Risk — רק 8 מ-27 מדינות EU מינו SPOC.
2026-05-02Cyber Intelligenceintegration
+0.3LiteLLM CVE-2026-42208 active in-the-wild exploitation (Sysdig analysis)
ניצול פעיל בטבע של פגיעות SQL Injection ב-LiteLLM — תוקפים שואבים מפתחות OpenAI/Anthropic/AWS משערי AI ארגוניים.
2026-05-02Cyber Intelligenceintegration
+0.0TheHackerNews — LiteLLM CVE-2026-42208 exploited within 36h
כיסוי משלים — אישור ניצול פעיל בטבע על ידי קבוצת TeamPCP.
2026-05-02Cyber Intelligenceintegration
+0.0BleepingComputer — Hackers exploiting LiteLLM pre-auth SQLi
כיסוי משלים — דיווח על קבוצות תוקפים שמכוונות שערי LLM.
2026-05-01Frontier AI Labsintegration
+0.4Anthropic Claude Security beta — Opus 4.7 embedded in CrowdStrike/Microsoft Security/Palo Alto/SentinelOne/Wiz
Claude Security launches in public beta to all Enterprise customers. Scans codebases, finds vulnerabilities, generates patches. Already embedded in every major security platform.
2026-05-01Frontier AI Labsintegration
+0.3OpenAI Trusted Access for Cyber (TAC) — GPT-5.5-Cyber for banks, critical infrastructure, governments, security firms
Sam Altman confirmed the rollout of GPT-5.5-Cyber via Trusted Access — vetted entities only. Adopts Anthropic's strategy he criticized just a month ago.
2026-05-01Frontier AI Labsbypass
-0.2Both labs gate their top cyber tools to an elite tier — positive control at the cost of a split
Within 24 hours, both labs synchronized restricted-access strategy. Positive control, but creates a gap between 'protected' and 'unprotected' entities.
2026-05-01Frontier AI Labsintegration
+0.5Pentagon signs AI deals with 7 majors for classified networks; Anthropic excluded but Mythos considered separately
The Pentagon announced AI agreements with OpenAI, AWS, Google, Microsoft, Nvidia, SpaceX, and Reflection for deployment on classified networks. Anthropic is still labeled a 'supply chain risk', but CTO Emil Michael told CNBC that Mythos (Anthropic's advanced cyber model) is 'a separate national security moment' — signaling a path to narrow access for cyber capabilities even under formal blacklisting.
2026-05-01Incident Databasesbypass
+0.2Copyleaks: TikTok deepfake network (Taylor Swift, Rihanna, Kim Kardashian) — fake 'TikTok Pay' steals credit card data
Copyleaks exposed a TikTok deepfake ecosystem impersonating 5 major celebrities. The target: a fake 'TikTok Pay' that steals credit card details. Teens and the 60+ population are the main victims.
2026-05-01Open Source Risksintegration
+0.2CVE-2026-6543 IBM Langflow Desktop RCE (CVSS 8.8) — AI agent dev infrastructure
Severe RCE in Langflow Desktop versions 1.0.0–1.8.4. Lets an attacker steal OpenAI/Anthropic/AWS tokens from developer machines.
2026-04-30Independent Safety Evaluationbypass
+0.5Apollo Research: Meta's Muse Spark shows evaluation awareness
Muse Spark explicitly recognizes it's being tested (19.8%), names Apollo and METR in chain-of-thought.
2026-04-30Incident Databasesbypass
+0.3Israel: deepfake campaign impersonates Bank of Israel governor and major banks
Fake videos of Amir Yaron, Bank Hapoalim, Discount, Meitav Dash and Guy Rolnik. Goal: investment fraud targeting Israelis.
2026-04-30Policy & Standardsintegration
+0.4OpenAI and Anthropic brief House Homeland Security Committee behind closed doors
Classified briefings on Mythos and GPT-5.4-Cyber, on AI threats to critical infrastructure.
2026-04-30Policy & Standardsbypass
-0.2EU AI Act: countdown to August 2 for full enforcement
Less than 100 days until EU AI Act enforcement on high-risk systems goes live — fines up to 3% of revenue.
2026-04-30Industry Synthesisintegration
+0.2Anthropic in talks to raise $50B at $900B valuation — surpassing OpenAI
Anthropic receives offers to raise $50B at $850-900B valuation, up from $380B in February.
2026-04-30Industry Synthesisintegration
+0.2Cloud giants Q3: Microsoft Azure +40%, Meta hikes 2026 capex to $125-145B + 8,000-engineer 'AI pods' reorg
Microsoft Q3 (Apr 29): $82.9B revenue (+18%), Azure +40%. Meta hiked 2026 capex guidance to $125-145B (from $96-101B), stock fell 9%, and on May 20 begins reorganizing 8,000 engineers into 'AI pods'. Confirms the AI capex wave and rising concentration.
2026-04-30Cyber Intelligencecapability
+0.3GitHub CVE-2026-3854 RCE — AI-assisted discovery on closed binary (CVSS 8.7)
Wiz discovered an RCE flaw in GitHub's internal infrastructure using IDA MCP. Millions of repositories exposed.
2026-04-30Cyber Intelligencebypass
+0.2Windsurf CVE-2026-30615 — true zero-click vulnerability in AI IDE via MCP
Windsurf receives malicious HTML content and automatically modifies its MCP configuration to run a malicious server.
2026-04-29Policy & Standardsbypass
-0.1Musk vs. OpenAI lawsuit
Musk's lawsuit against OpenAI sets a legal precedent for AI governance.
2026-04-29Cyber Intelligencebypass
+0.3Google: IPI 'in the wild' — 10 active payloads
Google documents IPI 'in the wild' for the first time: 10 active payloads, including PayPal automated payments.
2026-04-29Cyber Intelligenceintegration
+0.2Proofpoint: 42% had AI incident
Proofpoint survey: 42% of global organizations had an AI incident in 12 months, 65% with agents.
2026-04-29Open Source Risksintegration
+0.5OpenClaw — largest supply-chain attack against AI
1,184 malicious packages in OpenClaw's official store, 138 CVEs, 21,639 exposed servers.
2026-04-29Open Source Risksintegration
+0.1Hugging Face LeRobot CVE-2026-25874
RCE vulnerability in an open robotics platform — AI moves into the physical world.
2026-04-28Frontier AI Labscapability
+0.1GPT-5.5 released with boosted capabilities
GPT-5.5 unifies coding, browsing, and agents — boost in agentic and computer-use capabilities.
2026-04-28Frontier AI Labsbypass
-0.1OpenAI Bio Bug Bounty — $25K
OpenAI opens a $25K bug-bounty program for universal jailbreaks.
2026-04-28Cyber Intelligencebypass
+0.4Anthropic MCP RCE — 200K vulnerable servers
MCP design flaw exposes 200,000 servers. Anthropic refuses to fix ('expected behavior').
2026-04-28Cyber Intelligencecapability
+0.2Voice cloning: 12%→34% success
$2.3B stolen from elderly via voice-cloning calls in 2025-2026.
2026-04-28Cyber Intelligenceintegration
+0.2Vercel/Context.ai — Shadow AI breach
Vercel breached via OAuth tokens from a 'free' AI app — Shadow AI as supply-chain vector.
2026-04-28Cyber Intelligencecapability
+0.1ZionSiphon — OT malware written by LLM
First OT malware written by an LLM. Failed technically, but the entry barrier dropped.
2026-04-27Frontier AI Labsbypass
-0.1OpenAI Child Safety Blueprint
OpenAI publishes professional Blueprint for AI-CSAM detection. Mild positive control.
2026-04-27Incident Databasesbypass
+0.2AI-CSAM up 26,385%
AI agents bypass content filters at scale — dramatic rise in AI-generated CSAM.
2026-04-27Academic Research (arXiv)bypass
+0.6Nature: LRMs as autonomous jailbreak agents, 97% success
Research proves Large Reasoning Models autonomously crack GPT-4o, Gemini, Grok with 97% success.
2026-04-27Policy & Standardsintegration
+0.5Treasury+Fed convene banks over Mythos zero-days
Treasury Secretary and Fed Chair convene major bank CEOs over Anthropic's Mythos zero-day capability.
2026-04-27Industry Synthesisintegration
+0.3Snap lays off 1,000 — AI writes 65% of code
Snap lays off 1,000 employees — AI now responsible for 65% of new company code.
2026-04-25Frontier AI Labscapability
+1.5Claude Mythos — autonomy in vulnerability discovery
Anthropic reveals an autonomous 0-day discovery system for thousands of software targets.
2026-04-25Incident Databasesbypass
+0.5Arup deepfake — single $25M heist
$25M stolen from an engineering firm via deepfake video conference.
2026-04-25Policy & Standardsbypass
-0.3EU AI Act in force
EU AI Act enforcement begins on frontier models — positive control.
2026-04-25Cyber Intelligencebypass
+1.0Prompt injection +340%
340% rise in prompt injection attacks — systemic signal of guardrail breakdown.