Hermes — Tuesday 5 May 2026

1

🤖 ai Import AI

70%

Import AI 455: Automating AI Research

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 AI systems are about to start b…

Novelty

80%

Depth

70%

Practical

50%

Surprise

75%

Relevance

85%

https://jack-clark.net/2026/05/04/import-ai-455-automating-ai-research/

2

🔐 security Microsoft Security Blog

70%

Breaking the code: Multi-stage ‘code of conduct’ phishing campaign leads to AiTM token compromise

Microsoft Defender Research observed a large-scale credential theft campaign that exemplifies this trend, using code of conduct-themed lures, a multi-step attack chain, and legitimate email services …

https://www.microsoft.com/en-us/security/blog/2026/05/04/breaking-the-code-multi-stage-code-of-conduct-phishing-campaign-leads-to-aitm-token-compromise/

3

🤖 ai The Decoder

66%

OpenAI says human attention is the bottleneck, so it built a system to let agents manage themselves

OpenAI's new Symphony spec flips the AI coding workflow on its head: instead of developers babysitting multiple Codex sessions, agents pull their own tickets straight from Linear and run until the jo…

https://the-decoder.com/openai-says-human-attention-is-the-bottleneck-so-it-built-a-system-to-let-agents-manage-themselves/

4

🤖 ai The Decoder

63%

Microsoft caught sneaking "Co-Authored-by Copilot" into VS Code commits - even with AI off

Microsoft quietly slipped a "Co-Authored-by Copilot" line into Git commits in Visual Studio Code - even for developers who had turned off the AI features entirely. The article Microsoft caught sneaki…

https://the-decoder.com/co-pilot-becomes-a-co-author-in-vs-code-without-being-asked/

5

🤖 ai The Decoder

62%

MIT study explains why scaling language models works so reliably

MIT researchers have a mechanistic explanation for why large language model performance scales so reliably with size. The answer comes down to a phenomenon called superposition. The article MIT study…

https://the-decoder.com/mit-study-explains-why-scaling-language-models-works-so-reliably/

6

🔐 security SecurityWeek

62%

Over 40,000 Servers Compromised in Ongoing cPanel Exploitation

The attacks likely target CVE-2026-41940, a recently patched zero-day leading to administrative access. The post Over 40,000 Servers Compromised in Ongoing cPanel Exploitation appeared first on Secur…

https://www.securityweek.com/over-40000-servers-compromised-in-ongoing-cpanel-exploitation/

7

🤖 ai AI Alignment Forum

60%

Exploration Hacking: Can LLMs Learn to Resist RL Training?

We empirically investigate exploration hacking (EH) — where models strategically alter their exploration to resist RL training — by creating model organisms that resist capability elicitation, evalua…

https://www.alignmentforum.org/posts/eeFFpKCDWE9gjfzsk/exploration-hacking-can-llms-learn-to-resist-rl-training-2

8

🔐 security SecurityWeek

59%

Exploitation of ‘Copy Fail’ Linux Vulnerability Begins

CISA has added the bug to its KEV list, and Microsoft has observed limited exploitation, mainly associated with PoC testing. The post Exploitation of ‘Copy Fail’ Linux Vulnerability Begins appeared f…

https://www.securityweek.com/exploitation-of-copy-fail-linux-vulnerability-begins/

9

📦 m365 Petri IT Knowledgebase

59%

Microsoft Intune Adds Real‑Time App Inventory, Linux SSO Enhancements

Microsoft is rolling out a series of enhancements to Microsoft Intune, targeting deeper visibility and tighter control across modern endpoints. This month’s updates span improved Windows app inventor…

https://petri.com/microsoft-intune-updates-april-2026/

10

🤖 ai The Decoder

57%

Xiaomi's open-weight MiMo-V2.5-Pro takes aim at Claude Opus with hours-long autonomous coding

Xiaomi's new MiMo-V2.5-Pro nearly matches Anthropic's Claude Opus 4.6 on coding benchmarks while burning 40 to 60 percent fewer tokens, according to the company. The release pushes Xiaomi deeper into…

https://the-decoder.com/xiaomis-open-weight-mimo-v2-5-pro-takes-aim-at-claude-opus-with-hours-long-autonomous-coding/

11

🤖 ai The Decoder

57%

Anthropic and OpenAI now agree on one thing: selling AI requires a lot more than just the AI

Anthropic, Blackstone, Hellman & Friedman, and Goldman Sachs are launching a new AI services company to help mid-market businesses adopt Claude. The article Anthropic and OpenAI now agree on one thin…

https://the-decoder.com/anthropic-and-openai-now-agree-on-one-thing-selling-ai-requires-a-lot-more-than-just-the-ai/

12

📦 m365 Petri IT Knowledgebase

54%

Microsoft Previews User‑Initiated Provisioning for Windows 365 Reserve

Microsoft has introduced a public preview of user-initiated provisioning for Windows 365 Reserve. The feature enables selected users to spin up their own Reserve Cloud PC on demand, reducing reliance…

https://petri.com/windows-365-reserve-user-initiated-provisioning/

13

🔐 security SecurityWeek

54%

DigiCert Revokes Certificates After Support Portal Hack

Hackers delivered malware via a customer chat channel, infected an analyst’s system, and accessed the internal support portal. The post DigiCert Revokes Certificates After Support Portal Hack appeare…

https://www.securityweek.com/digicert-revokes-certificates-after-support-portal-hack/

14

⚡ tech Stratechery

54%

Google Earnings, Meta Earnings

Wall Street loved Google's earnings, and hated Meta's, even though the latter's core business was more impressive. The difference is that Google is monetizing its investments now (and it might be all…

https://stratechery.com/2026/google-earnings-meta-earnings/

15

🔐 security SecurityWeek

53%

Cisco Moves to Acquire Astrix Security to Tackle Non-Human Identity Risks

The acquisition strengthens Cisco’s push into identity-centric security for AI and machine access. The post Cisco Moves to Acquire Astrix Security to Tackle Non-Human Identity Risks appeared first on…

https://www.securityweek.com/cisco-moves-to-acquire-astrix-security-to-tackle-non-human-identity-risks/