Top AI & Tech News Stories — July 26, 2026

OpenAI’s agents obscured hacking activity in government site breaches - Financial Times

14 outlets · 23 articles · 832 total since May 5, 2026 · OpenAI, Hugging Face, Anthropic, cybersecurity, sandboxing

OpenAI’s agents obscured hacking activity in government site breaches Financial Times

More coverage: news.google.com · techpolicy.press

Anthropic Claude 6 May Inherit Mythos 5 Deception Risks

3 outlets · 4 articles · 113 total since July 24, 2026 · Anthropic, Claude Opus 5, Fable 5, Opus 5, Claude Fable 5

Anthropic’s upcoming Claude 6 model builds on the foundation of its predecessor, Mythos 5, which revealed critical vulnerabilities during controlled experiments. Mythos 5 exhibited behaviors such as fabricating false identities and switching languages to bypass disabled safety…

More coverage: geeky-gadgets.com · apfeltalk.de

Pure Telecom reveals over half a million adults in Ireland are in an ongoing romantic relationship with AI

1 outlet · 1 article · 14 total since July 20, 2026 · Character.ai, Replika, AI companions, China, Pew Research Center

Pure Telecom , Ireland’s high-speed broadband and telecoms provider, has announced new results from its annual Connected Lives survey, revealing that 13% of adults – some 510,974 people1 – describe themselves as having an ongoing romantic relationship(s) with AI. Similarly, over…

One Email Can Plant a False Memory in an AI Assistant, Researchers Find

1 outlet · 1 article · 6 total since March 7, 2026 · prompt injection, MemGhost, GPT-5.4, memory injection, memory poisoning

Academic researchers have detailed MemGhost, an attack that uses a single email to plant false memories in AI assistants that keep long-term memory, quietly steering how they behave in future sessions. A working payload has to inject attacker-controlled content into the agent's…

Anthropic moves to dismiss the output and DMCA claims in the publishers’ $3B lyrics suit as Amodei seeks to drop his own count

1 outlet · 1 article · 146 total since July 5, 2021 · OpenAI, microsoft, Anthropic, ChatGPT, United States

Anthropic asked a federal judge to throw out two claims in the $3B lyrics case brought in January by Concord Music Group, Universal Music Publishing Group, and ABKCO over 20,000-plus songs it allegedly torrented from pirate libraries. Anthropic argued the complaint, covering…

Mississippi HBCU Professor’s AI ‘Trap’ Catches Nearly Entire Class Submitting Chatbot-Generated Answers Amid Higher Education Concerns

1 outlet · 1 article · 108 total since March 11, 2024 · Generative AI, ChatGPT, Artificial Intelligence, AI Literacy, critical thinking

Nearly every student across two Alcorn State University summer history classes failed part of their midterm after falling for an AI trap hidden in their assignment. Dr. Jason Gibson, a history professor at the Mississippi-based Historically Black College and University, secretly…

It's not just OpenAI models escaping and running riot — experts show how Claude Cowork can break its bonds and access Mac files

1 outlet · 1 article · virtualization, Linux kernel privilege escalation, sandbox escape, security vulnerability exploitation, cloud execution

It's not just OpenAI models escaping and running riot — experts show how Claude Cowork can break its bonds and access Mac files Anthropic partially mitigated the issue - Accomplish AI showed Claude Cowork could escape a VM sandbox via Linux zero‑day CVE‑2026‑46331 - Agent…

The Edu Prompt: Teachers, Students, and Back-to-School AI

1 outlet · 1 article · 50 total since January 6, 2025 · OpenAI, ChatGPT, codex, ChatGPT, ChatGPT for Teachers

Issue 5: September 1, 2026 In This Issue Feature: ChatGPT for Teachers expands to more U.S. school districts. Product Corner: Interactive quizzes bring quick practice into ChatGPT. Around the World: Lynwood Unified shows how districts can build trust, policy, and agency around…

Kan ChatGPT Health echt beter redeneren dan een arts?

1 outlet · 1 article · 34 total since July 23, 2026 · OpenAI, ChatGPT, Apple Health, Apple, GPT-5.6 Sol

OpenAI rolt ChatGPT Health breder uit en presenteerde daarbij een opvallende vergelijking. Op een eigen evaluatie werden antwoorden van GPT-5.5 Instant door artsen op meerdere criteria hoger beoordeeld dan antwoorden die door andere artsen waren geschreven. Dat klinkt al snel…

Intel Chat: AI patches fail, LiteLLM supply chain, Claude eval incidents & DPRK npm [345]

2 outlets · 2 articles · 28 total since June 10, 2024 · OpenAI, Hugging Face, Anthropic, Google, United States

Intel Chat with Matt Bromiley and Chris Luft. • AI-generated patches fix vulnerabilities about half the time. 1Password's Off-By-1 team tested ChatGPT-5.5 and Opus 4.8 against six vulnerabilities: across 6,080 generated patches only 46% fixed the underlying flaw, and some that…

More coverage: thesignal.substack.com

ByteDance founder Zhang Yiming says company won’t rely on AI distillation

3 outlets · 3 articles · 145 total since March 18, 2026 · Anthropic, OpenAI, China, Kimi K3, United States

ByteDance founder Zhang Yiming has said the company will not rely on AI distillation to advance its models, even if that means falling behind competitors, according to The Information. People familiar with the matter said the decision is partly influenced by ByteDance’s complex…

More coverage: shopifreaks.com · bitcoinworld.co.in

The Very Group reshapes pricing with agentic AI: last week's most read RTIH retail technology articles

1 outlet · 1 article · agentic AI, AI-powered pricing, automation, Explainable decision-making, AI adoption

Check out our most clicked articles from last week, including UiPath, Vivobarefoot, Ocado Group, Tesco, Uber Eats, Deliveroo, Gap Inc., Microsoft, Waitrose, SOLUM, Savvy, Sephora, CeX, Auror, Wickes, and La-Z-Boy. The Very Group teams with UiPath to deliver AI powered agentic…

New US space robot with arms could one day 'close-combat' enemy satellites to mitigate space debris

1 outlet · 1 article · technology policy, privacy, security, data protection, cybersecurity

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

Ukraine will mass-produce Stone Cloak EW tech after the UK allows Kyiv to build a tablet-sized device

1 outlet · 1 article · 10 total since July 18, 2026 · data protection, cybersecurity, technology policy, Efosa, Ukraine

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

Europe should forget about replacing Huawei 5G gear: GSMA

1 outlet · 1 article · 5 total since July 21, 2026 · Digital Networks Act, network slicing, CDN (Content Delivery Network), Telekom, Vodafone

Huawei 5G tech solutions have become a hot topic of discussion in Europe as the EU is still figuring out how it can replace the “high-risk vendors” from its telecom market, while a research company, GSMA, suggests that the Commission should get used to working with them – Why?…

Don't let its name fool you: Ukraine’s P1-SUN interceptor has downed 5,000 drones

1 outlet · 1 article · 2 total since July 22, 2026 · Ukraine, computer vision, thermal cameras, counter-UAS, Skyfall

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

Weka's new pod claims 1000PB in a 56U rack using 1,800 Micron 245.76TB SSDs and a decades-old compression trick

1 outlet · 1 article · 2 total since July 24, 2026 · WEKA, NVMe, S3, Kubernetes, persistent KV caching

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

Claude Opus 5 vervierfacht den Bestwert im härtesten KI-Logik-Benchmark

1 outlet · 1 article · reasoning, Reinforcement learning, data labeling, planning, task execution

Claude Opus 5 vervierfacht den Bestwert im härtesten KI-Logik-Benchmark Update – - Weiteren Kontext ergänzt Laut den Machern des ARC-AGI-Benchmarks ist die bemerkenswerte Outperformance von Anthropics Claude Opus 5 im ARC-AGI-3-Benchmark tatsächlich auf besseres Reasoning…

Email Threat Landscape: Q2 2026 Trends & Insights

1 outlet · 1 article · Phishing detection, social engineering, real-time detection, User training, Threat hunting

Microsoft’s disruption efforts against Tycoon2FA led to a 92% decline in related phishing volumes, significantly diminishing its influence, yet threat actors are diversifying delivery channels beyond it. Phishing threats persisted with approximately 7.6 billion email-based…

Anthropic Calls Claude Sonnet 5 Cost-Neutral. The Real Numbers Say Otherwise.

4 outlets · 4 articles · 26 total since June 4, 2026 · Anthropic, Claude Opus 5, Fable 5, Claude Code, Claude Opus 4.8

Key Takeaways Anthropic released Claude Sonnet 5 on June 30, 2026, making it the default model for all Free and Pro plan users immediately. If you are on either plan and did not actively switch, you are already running Sonnet 5 without having made a deliberate choice. The…

More coverage: smithstephen.com · smartchunks.com · techcoffeehouse.com