Hermes — Tuesday 7 April 2026

1

🔐 security Microsoft Security Blog

75%

Inside an AI‑enabled device code phishing campaign

A new wave of device code phishing shows how threat actors are scaling account compromise using AI and end‑to‑end automation. This campaign goes beyond traditional phishing by generating live authent…

Novelty

75%

Depth

70%

Practical

75%

Surprise

65%

Relevance

90%

https://www.microsoft.com/en-us/security/blog/2026/04/06/ai-enabled-device-code-phishing-campaign-april-2026/

2

🤖 ai The Decoder

71%

AI offensive cyber capabilities are doubling every six months, safety researchers find

AI models are rapidly improving at exploiting security vulnerabilities. According to a new study, their offensive cyber capability has been doubling every 5.7 months since 2024, with Opus 4.6 and GPT…

Novelty

80%

Depth

60%

Practical

65%

Surprise

75%

Relevance

85%

https://the-decoder.com/ai-offensive-cyber-capabilities-are-doubling-every-six-months-safety-researchers-find/

3

🔐 security SecurityWeek

67%

Google DeepMind Researchers Map Web Attacks Against AI Agents

A vulnerability named ‘AI Agent Traps’ allows attackers to manipulate, deceive, and exploit visiting agents via malicious web content. The post Google DeepMind Researchers Map Web Attacks Against AI …

https://www.securityweek.com/google-deepmind-researchers-map-web-attacks-against-ai-agents/

4

📦 m365 Petri IT Knowledgebase

67%

Why Active Directory Password Policy Fails Modern Attacks (and What Admins Need Instead)

According to Microsoft’s latest Digital Defense Report, 97% of identity attacks are password spray attacks. This statistic shows that most attackers aren’t using sophisticated cracking techniques. In…

https://petri.com/why-ad-password-policy-fails-modern-attacks/

5

🤖 ai The Decoder

66%

Sycophantic AI chatbots can break even ideal rational thinkers, researchers formally prove

A new study by researchers from MIT and the University of Washington shows that even perfectly rational users can be drawn into dangerous delusional spirals by flattering AI chatbots. Fact-checking b…

https://the-decoder.com/sycophantic-ai-chatbots-can-break-even-ideal-rational-thinkers-researchers-formally-prove/

6

🔐 security Microsoft Security Blog

66%

Storm-1175 focuses gaze on vulnerable web-facing assets in high-tempo Medusa ransomware operations

The financially motivated cybercriminal threat actor Storm-1175 operates high-velocity ransomware campaigns that weaponize recently disclosed vulnerabilities to obtain initial access, exfiltrate data…

https://www.microsoft.com/en-us/security/blog/2026/04/06/storm-1175-focuses-gaze-on-vulnerable-web-facing-assets-in-high-tempo-medusa-ransomware-operations/

7

⚡ tech The Verge

64%

Iran threatens OpenAI’s Stargate data center in Abu Dhabi

An October 2025 image of OpenAI’s UAE Stargate data center under construction. | Image: G42 Iran's Islamic Revolutionary Guard Corps (IRGC) has published a video threatening OpenAI's planned Abu Dhab…

https://www.theverge.com/ai-artificial-intelligence/907427/iran-openai-stargate-datacenter-uae-abu-dhabi-threat

8

🤖 ai AI Alignment Forum

64%

AIs can now often do massive easy-to-verify SWE tasks and I've updated towards shorter timelines

I've recently updated towards substantially shorter AI timelines and much faster progress in some areas. [1] The largest updates I've made are (1) an almost 2x higher probability of full AI R&D autom…

https://www.alignmentforum.org/posts/dKpC6wHFqDrGZwnah/ais-can-now-often-do-massive-easy-to-verify-swe-tasks-and-i

9

📦 m365 Microsoft 365 Blog

63%

Copilot Cowork: Now available in Frontier

Today, Copilot Cowork—designed for long-running, multi-step work in Microsoft 365—is available via the Frontier program. The post Copilot Cowork: Now available in Frontier appeared first on Microsoft…

https://www.microsoft.com/en-us/microsoft-365/blog/2026/03/30/copilot-cowork-now-available-in-frontier/

10

🤖 ai Import AI

62%

Import AI 452: Scaling laws for cyberwar; rising tides of AI automation; and a puzzle over gDP forecasting

Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Uh oh, there’s a scaling war fo…

https://jack-clark.net/2026/04/06/import-ai-452-scaling-laws-for-cyberwar-rising-tides-of-ai-automation-and-a-puzzle-over-gdp-forecasting/

11

🔐 security Krebs on Security

62%

Germany Doxes “UNKN,” Head of RU Ransomware Gangs REvil, GandCrab

An elusive hacker who went by the handle "UNKN" and ran the early Russian ransomware groups GandCrab and REvil now has a name and a face. Authorities in Germany say 31-year-old Russian Daniil Maksimo…

https://krebsonsecurity.com/2026/04/germany-doxes-unkn-head-of-ru-ransomware-gangs-revil-gandcrab/

12

📦 m365 Microsoft 365 Blog

60%

New and improved: Multi-agent orchestration, connected experiences, and faster prompt iteration

Learn what’s new in Copilot Studio: Multi-agent systems are now generally available, plus recent updates to the Prompt Editor and governance controls. The post New and improved: Multi-agent orchestra…

https://www.microsoft.com/en-us/microsoft-copilot/blog/copilot-studio/new-and-improved-multi-agent-orchestration-connected-experiences-and-faster-prompt-iteration/

13

🤖 ai The Decoder

58%

Alibaba's Qwen team makes AI models think deeper with new algorithm

Reinforcement learning hits a wall with reasoning models because every token gets the same reward. A new algorithm from Alibaba's Qwen team fixes this by weighting each step based on how much it shap…

https://the-decoder.com/alibabas-qwen-team-makes-ai-models-think-deeper-with-new-algorithm/

14

🤖 ai The Decoder

57%

Alibaba's Qwen team built HopChain to fix how AI vision models fall apart during multi-step reasoning

When AI models reason about images, small perceptual errors compound across multiple steps and produce wrong answers. Alibaba's HopChain framework tackles this by generating multi-stage image questio…

https://the-decoder.com/alibabas-qwen-team-built-hopchain-to-fix-how-ai-vision-models-fall-apart-during-multi-step-reasoning/

15

🔐 security SecurityWeek

56%

Fortinet Rushes Emergency Fixes for Exploited Zero-Day

The improper access control bug in FortiClient EMS allows unauthenticated attackers to execute arbitrary code remotely. The post Fortinet Rushes Emergency Fixes for Exploited Zero-Day appeared first …

https://www.securityweek.com/fortinet-rushes-emergency-fixes-for-exploited-zero-day/