
OpenAI Launches GPT-5.6 Models Sol, Terra and Luna for Broad Testing
OpenAI expands testing of GPT-5.6 models Sol, Terra and Luna after meetings with US Commerce Department officials, marking a significant step in frontier AI deployment.
The most recent stories from CapisTech, covering AI, cybersecurity, startups, gadgets, software and innovation.

OpenAI expands testing of GPT-5.6 models Sol, Terra and Luna after meetings with US Commerce Department officials, marking a significant step in frontier AI deployment.

Google advances its agentic AI strategy with Gemini 3.5 and Gemini Omni, positioning AI as proactive systems that plan, reason and act across work tools rather than passive chatbots.

Meta launches Muse Image, its most advanced in-house image generation model, alongside an early preview of Muse Video with native audio support and content verification features.

SpaceXAI launches Grok 4.5 as a business-focused AI model for coding, engineering and agentic tasks, priced competitively at $2 input and $6 output per million tokens.

DeepSeek is reportedly building its own AI inference chip to reduce reliance on Nvidia and Huawei hardware, facing significant obstacles including manufacturing restrictions.

MIT researchers develop Murakkab system to improve speed and reduce energy consumption in multistep AI agent workflows, potentially reshaping agent economics.

Researchers demonstrate GhostCommit, an attack that manipulates AI coding agents through hidden instructions embedded in images, raising security concerns for AI-assisted development.

Prime Intellect secures $130M Series A at $1B valuation to offer enterprises a full-stack platform for training and refining their own AI agents without relying on frontier models.

Cloud security firm Sysdig reports JadePuffer, the 1st documented ransomware attack executed entirely by an LLM agent without human intervention.

The US Department of Homeland Security confirms a cyber breach of HSIN, an unclassified information sharing platform, with intrusion likely occurring between late May and early June 2026.

Microsoft's July 2026 Patch Tuesday addresses 569-570 vulnerabilities, including 3 zero-days actively exploited in the wild, requiring immediate patching.

Researchers use artificial intelligence to discover GhostLock, a Linux privilege escalation vulnerability that remained hidden for nearly 15 years.

A coordinated operation involving Google, the FBI and international partners takes down the NetNut residential proxy network compromising approximately 2 million devices.

SOCRadar links the massive FortiBleed campaign exposing credentials from over 430,000 Fortinet devices to INC Ransom and Lynx ransomware operations.

CISA adds CVE-2026-45659, a remote code execution flaw in Microsoft SharePoint Server, to its Known Exploited Vulnerabilities catalog requiring federal agencies to patch immediately.

A 16-year-old use-after-free bug in the Linux KVM hypervisor allows guest virtual machines to escape to the host, affecting both Intel and AMD systems.

The European Union and United Kingdom announce coordinated sanctions against Russian individuals and organizations linked to cyber operations following attacks on European energy infrastructure.

A new Android malware called RedWing, a variant of Oblivion, is being rented as a banking fraud service on Telegram with phone takeover and OTP capture capabilities.

SambaNova Systems closes a $1 billion Series F funding round at an $11 billion valuation, with JPMorgan Chase adopting its platform for private model inference.

Jeff Bezos' Blue Origin raises $10 billion in its 1st outside funding round at a $130 billion valuation, with Coatue expected to lead and Bezos contributing $2 billion.