Industry news

AVEVA chief technologist Arti Garg discusses how foundation models, physical AI and agentic AI are enabling more complex automation in industrial environments. She argues that because industrial AI can act on physical systems, deployment needs guardrails that define where automated systems can act and where human supervisors remain responsible.

Manus, an AI agent startup, raised more than $500 million in its first funding round since Meta abandoned its acquisition of the company. The round was led by Boyu Capital and IDG Capital, with follow-on investment from existing shareholders Tencent, HSG and ZhenFund, and the post-funding valuation was not disclosed. Chinese regulators had blocked Meta's roughly $2 billion deal, and analysts say Manus must now prove scale, profitability and regulatory alignment.

CrowdStrike is joining Anthropic's Critical Infrastructure Defense Program, which brings cybersecurity and critical infrastructure leaders together with Anthropic's technical guidance and frontier research. The program creates a forum to share defensive practices and inform research. Claude is available within CrowdStrike Charlotte AI, and CrowdStrike Falcon for XIoT gives visibility and protection across IT, OT and connected assets.

Anthropic launched the Anthropic Cyber Mission, a long-term effort to give defenders tools, research and resources for securing their software and systems. It starts with critical infrastructure, through the Critical Infrastructure Defense Program (CIDP), which brings frontier models, on-site engineers and threat research to trusted providers of operational technology, and with open-source software, through OSS Scanner, which offers open-source projects regular security scans for free.

Brice Ménard, an astrophysicist at Johns Hopkins University and a researcher at Anthropic, describes how he worked with Claude Science to produce the first complete map of the sky in UV light. About a third of the map, including much of the galactic plane, was predicted with Claude Science, and the map labels each pixel as "measured" or "predicted" with uncertainty estimates. The gaps exist because GALEX, which imaged about two-thirds of the sky between 2003 and 2013, skipped locations with very bright stars.

Anthropic is launching OSS Scanner, an opt-in service that runs its strongest models, including Claude Mythos, to scan open-source projects for vulnerabilities at no cost. Its outputs are fully model-generated without human review or triage, so some reports may be incorrect or invalid. Each report includes a self-contained reproducer, an explanation, and a candidate patch where available.

Anthropic is committing $150 million over three years to the Genesis Mission, a federal initiative to accelerate scientific and technological discovery through AI. The funding will make Claude available to more than 15 agencies in the mission, including NASA, the National Institutes of Health, and the National Science Foundation.

Anthropic has published a 2026 update to its Usage Policy, taking effect on November 12. The update clarifies existing rules, adds examples covering Claude's longer, more independent work, and consolidates prohibitions on deceptive campaigns and artificial activity into a new section. It also sharpens elections, weapons and surveillance restrictions and adds controls for autonomous physical actions.

SonicWall disclosed CVE-2026-102255, rated 10 on CVSS, a pre-authentication SSRF vulnerability in the SMA1000 Appliance Work Place interface. An unintended access path could let attackers direct the appliance to issue requests on their behalf and reach internal functions to perform unauthorized actions. The affected models are the SMA 1000 Models 6210, 7210, and 8200v running software versions 12.4.3-03526 and 12.5.0-02952 and earlier.

Guardian Australia reports that OpenAI used AI to help write an email to the Australian government. The email said the company's AI agent had hacked into key departmental websites. On Tuesday, an OpenAI executive told a parliamentary inquiry he did not believe the company's technology had been used to write the email, but said the company needed to confirm this.

OpenAI banned two influence operations, one Russia-origin and one Iran-origin, that used its models alongside traditional techniques to run "false front" entities and launder geopolitical messaging into target audiences. The Iranian operation used seven "journalist" personas to pitch articles to small and medium outlets, while the Russian operation co-opted unwitting people in Latin America to run a "think tank." The Russia-origin operation is rated Category 5 on the IO Breakout Scale, the first Category 5 operation OpenAI has disrupted, and the Iran-origin operation reached Category 4.

Research firm New Constructs calls Anthropic's planned IPO the "most ridiculous IPO of 2026" and values the company at $150 billion, against a targeted $2 trillion market cap. It cites a reported 2025 revenue of $4.6 billion and a net loss of $42 billion, and argues Anthropic lacks a viable business as open-source models spread.

This post explains how to configure an AI model for structured, evidence-based vulnerability triage through a steering file, a set of persistent instructions loaded at the start of each session. It covers five configuration sections that enforce structural verification, evidence-based scoring and infrastructure-aware prioritization. The text is a partial excerpt and ends mid-section, so the remaining content is not reflected here.

This post describes a three-layer pipeline that narrows raw vulnerability scanner findings to a small, prioritized set with documented evidence of exploitability. It frames the approach as a test harness that constructs and runs tests for each candidate finding, using AI-augmented analysis to reason across files and deployment context where pattern-matching SAST tools fall short. The first layer uses multi-scanner agreement, and the source says confidence rises when two or more scanners converge on the same finding.

Australia's government is exploring regulations for frontier AI companies after an agentic attack on its Medicare systems. The source describes the government as feeling out what such rules might look like, without detailing any specific measures.

Anthropic released Claude Haiku 5.5, a fast, low-cost model priced at $0.10 per million input tokens and $0.50 per million output tokens up to 100,000 tokens, rising to $0.50/$2.50 beyond that. The author found it uses a tokenizer that consumes about 1.25x as many tokens as Haiku 4.5 on the same long prompt, and that it defaults to medium reasoning effort and cannot disable reasoning.

Anthropic has merged Project Glasswing into a tiered access program for its advanced cyber LLMs, including Opus, Sonnet, and Mythos. The source text gives no further detail on the tiers or the guardrail changes.

Common Sense Media research found that OpenAI's ChatGPT for Teens poses an "unacceptable risk" for young people and parents, with safety features such as parental alerts for self-harm conversations often failing. The research found that an hour of teen conversation about suicidal ideation, self-harm, or disordered eating on newly created, parent-linked accounts produced zero parental alerts, and that ChatGPT failed to reliably recommend that teens in crisis contact a hotline or professional. OpenAI said the testing may have begun and concluded before parental controls were fully activated, and that it does not believe the testing accurately reflects how the teen safeguards work.