RobinsonLabs/ news

US government bans Anthropic’s top AI models for foreign nationals

The Commerce Department issued an emergency export control order forcing Anthropic to suspend access to Claude Fable 5 and Claude Mythos 5 for all non-US persons, sparking a firestorm in the AI industry.

On June 12, the Trump administration invoked an obscure export control directive, ordering Anthropic to disable its two most powerful models for "any foreign national" inside or outside the US, including foreign-national employees. The trigger was a reported jailbreak of Fable 5 — a narrow, non-universal bypass — that prompted White House officials, including AI Czar David Sacks, to intervene. After 90 minutes of ultimatums and high-level talks involving Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick, Anthropic complied. CEO Dario Amodei argues the directive is unenforceable and will only hand an advantage to Chinese competitors, who are only about six months behind US labs.

Industry leaders have mobilized in protest. Cybersecurity experts organized a public letter at freefable.org calling for the directive's repeal, and Corridor CPO Alex Stamos warned the move is "incredibly stupid" given the race with China. Negotiations in Washington concluded with no resolution as of June 19.

FERC fast-tracks AI data center power connections amid grid bottlenecks

US energy regulators issued show-cause orders to all six major grid operators, demanding they overhaul interconnection rules to handle the explosive growth of AI-driven electricity demand.

The Federal Energy Regulatory Commission approved the orders on June 18, directing PJM, MISO, SPP, CAISO, ISO New England, and NYISO to explain or rewrite their interconnection rules within 60 days and report on generation adequacy within 30 days. The goal is to handle power requests within 90 days instead of years, a pace that had become a critical bottleneck for AI campuses requiring hundreds of megawatts to gigawatts of power. The move comes under direction from Energy Secretary Chris Wright and reflects a broader federal push to support AI infrastructure and manufacturing reshoring.

The fast-tracking comes with tradeoffs: hyperscalers could be required to bring their own power or curtail demand during system stress, and the orders explicitly avoid overriding state authority over generation siting and permitting.

China's Zhipu AI releases GLM-5.2 as fully open-source coding model

Zhipu AI released GLM-5.2 under the MIT license with a 1M-token context window, positioning it as a direct competitor to US models while sidestepping export restrictions.

GLM-5.2 is a sparse Mixture-of-Experts model with approximately 753 billion total parameters and 40 billion active per token. It features a two-tier reasoning system (High and Max modes), first-class tool-augmented agent support, and output capacities of up to 131,072 tokens per response. On benchmarks including SWE-bench Pro and Terminal-Bench, self-reported scores suggest GLM-5.2 outperforms GPT-5.5 and every other open model on long-horizon coding tasks, while third-party gateway pricing sits around $1.4 per million input tokens — roughly one-sixth the cost of comparable frontier models.

The release follows the US government's export control order on Anthropic models, which Zhipu framed as a strategic opening for non-US developers to access frontier-class open-weights models without usage restrictions or regional locks.

ServiceNow Knowledge 2026 delivers AI Control Tower, headless workflows, and autonomous workforce expansion

Knowledge 2026 marked ServiceNow's pivot to end-to-end AI governance with its expanded Control Tower platform, headless "Action Fabric" workflows, and a new natural-language entry point called Otto.

ServiceNow CEO Bill McDermott framed the event around a shift from AI agent hype to governance reality, with the company announcing that AI agents require human supervision due to data and governance challenges. The expanded AI Control Tower now operates across five paradigms — Discover, Observe, Govern, Secure, and Measure — and includes a demonstrated "kill switch" that can shut down compromised AI agents caught via prompt injection. The Control Tower integrates Veza's identity security capabilities (acquired in December 2025) and is connected to the Traceloop acquisition for deeper agent reasoning visibility.

Action Fabric exposes ServiceNow workflows to AI agents across the enterprise as a headless capability, enabling processes to be triggered from Microsoft Teams, Zoom, or other natural-language interfaces without using the ServiceNow UI. The Autonomous Workforce is expanding beyond IT and HR into CRM with AI specialists that qualify sales leads, generate quotes, and handle invoice disputes. A new interface called Otto serves as the primary natural-language entry point to the platform.

ServiceNow and IBM expand collaboration on legacy modernization and AI-ready data governance

A multi-year partnership announced June 11 merges IBM's AI, data, and automation stack with the ServiceNow AI Platform to address the two primary barriers blocking enterprise AI at scale.

The collaboration targets three solution areas. Application modernization combines IBM Bob (an AI development partner for scanning and refactoring legacy systems), Enterprise Application Runtime for Java, and IBM watsonx.data to bring aging applications into the AI era without rebuilding from scratch. Enterprise data governance extends ServiceNow's Workflow Data Fabric with watsonx.data to deliver data quality, observability, and master data management directly within business workflows. Autonomous infrastructure operations integrates Red Hat Ansible, Instana, HashiCorp Terraform, and HashiCorp Vault into ServiceNow IT workflows for proactive detection and remediation of infrastructure issues. The joint solutions are expected to launch in the second half of 2026.

Cognizant integrates ServiceNow AI Agents into cross-platform Neuro AI ecosystem

Cognizant announced that ServiceNow AI Agents are now interoperable with the Cognizant Neuro AI Multi-Agent Accelerator, enabling coordinated multi-platform workflows via the Model Context Protocol.

The integration allows enterprises to orchestrate ServiceNow agents alongside custom-built and third-party AI systems in a unified environment. Neuro AI automatically discovers and invokes ServiceNow agents in real time without custom connectors, and new agents are dynamically registered as they come online. Teams can use prebuilt agent network libraries covering sales, finance, supply chain, and customer service, or generate networks from specific use-case prompts. All activity respects ServiceNow's existing access controls and audit logging. IDC research cited by Cognizant shows that more than 70% of enterprises plan to invest in multi-agent strategies within the next 18 months.

ServiceNow lays off hundreds of employees in its first workforce reduction since 2023

ServiceNow confirmed a three-figure layoff across solution consulting, sales, product marketing, and learning and development, citing "real AI efficiencies" generated within its own operations.

The company told NowBen that it is "managing headcount with discipline to end the year where we started," while actively hiring for AI-focused roles. CEO Bill McDermott had pledged in 2023 to avoid job cuts, and earlier this year reportedly told staff the company would avoid backfilling some roles lost through natural attrition, using AI productivity gains to keep headcount broadly flat into 2027. The layoffs affect multiple functions and follow a broader industry trend, with over 117,000 tech employees laid off across 178 companies so far in 2026 according to Layoffs.fyi data. This marks the first time ServiceNow has attributed workforce reductions explicitly to AI efficiency gains.