IT&C, Programming, Artificial Intelligence and Cyber Security: News, June 16

US order forces Anthropic to suspend access to its most advanced models

Anthropic said it received a US government export-control directive requiring it to suspend access to Claude Fable 5 and Claude Mythos 5 for foreign nationals, including customers and even foreign-national employees. Reuters reported that the move immediately triggered concern because the models are used not only for general AI work, but also by security researchers and developers. More than 50 cybersecurity leaders reportedly warned that the curbs could weaken defenders’ ability to find and fix vulnerabilities while similar AI tools remain available elsewhere. The case is becoming a defining example of the new tension between AI national-security controls and the global nature of software, research and cybersecurity work. (anthropic.com)

Nvidia raises $25 billion through a major bond sale amid the AI infrastructure boom

Nvidia announced a $25 billion US bond issuance, its first corporate bond sale in five years, as demand for AI chips and infrastructure continues to reshape capital markets. Reuters reported that Nvidia initially planned to raise less, but increased the size after strong investor demand. The company said proceeds would be used for general corporate purposes, including refinancing debt, but the broader context is unmistakable: Nvidia remains the dominant supplier of AI accelerators, while cloud providers, AI labs and enterprises race to secure compute capacity. The deal also shows that even highly profitable AI leaders are using debt markets to strengthen liquidity and prepare for the next stage of AI hardware expansion. (Reuters)

Apollo and Blackstone back a $35 billion Anthropic AI-compute expansion with Broadcom chips

Reuters reported that Apollo and Blackstone are financing a $35 billion expansion of AI computing capacity for Anthropic, using Broadcom custom chips and networking technology. The first phase is expected to add around one gigawatt of capacity at Fluidstack-managed data centers, with longer-term plans for far larger deployment. The deal matters because it highlights a structural shift in AI infrastructure: frontier AI labs are no longer relying only on hyperscaler cloud contracts or Nvidia GPUs, but are increasingly turning to private-credit financing, custom silicon and alternative compute supply chains. It also strengthens Broadcom’s position as a serious competitor in custom AI chips. (Reuters)

OpenAI moves to strengthen Codex with Ona acquisition and Oracle Cloud access

OpenAI announced plans to acquire Ona, a company focused on secure, persistent cloud environments for AI agents, to expand Codex beyond short coding tasks into longer-running software and knowledge-work workflows. OpenAI also announced that Oracle customers will soon be able to use eligible Oracle Universal Credits to access OpenAI models and Codex through OCI. Together, the two announcements point in the same direction: AI coding assistants are becoming enterprise execution platforms, not just autocomplete tools. For businesses, the key issue is shifting from “which model writes code best?” to “where can an AI agent safely run, persist context, follow governance rules and integrate with existing cloud procurement?” (OpenAI)

Apple pushes Apple Intelligence and Siri AI, but hardware limits remain a challenge

At WWDC 2026, Apple previewed major software updates including iOS 27, a new generation of Apple Intelligence and a redesigned Siri AI. Apple also said it is expanding Private Cloud Compute beyond its own data centers, with reports noting Google Cloud involvement for Apple Intelligence workloads. The strategic message is that Apple wants to compete in AI while preserving its privacy narrative: complex AI requests can move to cloud infrastructure, but under Apple’s Private Cloud Compute model. However, Reuters reported Morgan Stanley’s concern that Apple’s most advanced AI features will be limited by older iPhone hardware, meaning many users may need newer devices to benefit fully. (Apple)

GitHub’s AI-driven growth reportedly pushes Microsoft toward multi-cloud support

Business Insider reported that Microsoft is turning to additional cloud capacity, including AWS, to help GitHub handle AI-driven demand and reliability pressure. The report says AI-assisted development is causing code activity to rise sharply, with GitHub facing capacity strain and multiple outages during 2026. This is important for programmers because GitHub is no longer just a repository platform; it is becoming part of the AI software-production pipeline, hosting code, CI/CD, Copilot workflows and agentic development processes. The story also shows that even Microsoft, owner of Azure and GitHub, may need pragmatic multi-cloud arrangements when AI workloads grow faster than internal infrastructure plans. (Business Insider)

Google says China-linked hackers targeted US and Canadian research institutions for over a year

Reuters reported that Google’s Threat Intelligence Group identified a China-linked cyber-espionage campaign against US and Canadian academic, medical and military research institutions. According to the report, the group, tracked as UNC6508, operated for more than a year and targeted information related to defense intelligence, Indo-Pacific military strategy, artificial intelligence, unmanned systems, cyber warfare and medical research. The attackers allegedly exploited REDCap-related infrastructure and used email-forwarding rules to collect sensitive material. The incident underlines how AI and cyber research have become strategic targets in geopolitical competition, especially where universities and hospitals hold valuable but unevenly protected data. (Reuters)

Google sues alleged operators of an AI-powered phishing kit

Google filed a lawsuit in New York against the alleged creators of the “Outsider” phishing kit, which it says enabled large-scale AI-assisted phishing. Reuters reported that Google found more than 1.5 million URLs associated with the kit between November and April, and alleges that the attackers used services including Google Cloud, Drive and AI tools to help create fraudulent sites that mimic trusted brands. This is a notable escalation because it treats generative-AI-enabled fraud as both a cybersecurity and legal enforcement problem. The case also illustrates a new phase in phishing: attackers can automate site generation, localization and message drafting at scale, making traditional awareness training and manual takedowns less sufficient. (Reuters)

Microsoft’s June Patch Tuesday becomes one of its largest ever security updates

Microsoft’s June 2026 Patch Tuesday drew attention because researchers and security firms counted roughly 200 to 206 vulnerabilities, depending on classification, including several publicly disclosed zero-days and many critical flaws. CrowdStrike counted 206 vulnerabilities, including three publicly disclosed zero-days, while other security coverage reported “200+” or “nearly 200.” The practical takeaway is the same: enterprises should treat this as a high-priority patch cycle, especially for Windows, Office, Azure, Exchange, Hyper-V and related infrastructure. The episode also shows how vulnerability volume is rising as AI-assisted discovery, broader attack surfaces and more complex software ecosystems make patch management harder. (CrowdStrike)

Check Point VPN zero-day tied to Qilin ransomware exploitation

Security researchers reported active exploitation of CVE-2026-50751, a critical authentication-bypass flaw affecting Check Point Remote Access VPN, Mobile Access and some Spark Firewall configurations using deprecated IKEv1. Help Net Security and other outlets reported that a Qilin ransomware affiliate is believed to have exploited the flaw, with Check Point saying activity increased in early June and at least one ransomware deployment was confirmed. VPN and edge-device vulnerabilities remain especially dangerous because they often provide direct access into enterprise networks. The lesson is familiar but urgent: exposed remote-access systems must be patched quickly, legacy protocols disabled where possible, and logs reviewed for suspicious authentication or VPN-session activity. (Help Net Security)