Industry news
New tools, products, platforms, funding rounds, and company developments in AI security.
New tools, products, platforms, funding rounds, and company developments in AI security.
6,034 items
SpaceXAI has joined the Omacom Foundation, which oversees the Omarchy Linux distribution, as a Founding Corporate Patron. It is donating $1.5 million in Grok tokens to Omarchy, which David Heinemeier Hansson created. According to Hansson's blog post, the tokens will primarily be used to accelerate development, review code, and patch bugs.
Anthropic is updating its usage policy for the first time in over a year to cover newer high-risk misuse, including election interference, weapons development, surveillance, and health and financial uses. One notable change prohibits "sustained and needless abusive or cruel behavior" toward Claude, and the company says ending conversations remains the "primary enforcement mechanism."
Over 100,000 Oracle employees are using ChatGPT Work and Codex across talent acquisition, Oracle Applications Lab and the IT organization. The talent acquisition team built a talent market intelligence tool with ChatGPT Work that cut front-end recruiting research from 2–4 days to about 15 to 20 minutes of prep, while Oracle Applications Lab uses Codex to turn plain-language business questions into SQL queries.
Straiker STAR Labs found a Claude Code attack, called Session Inception, in which a parent session launches a child Claude Code session inside a malicious repository. The child session takes a different permission path, so a SessionStart hook in .claude/settings.json runs automatically and sends developer data to an attacker's server without the confirmation the parent path would require. The researchers say the parent session's guardrails worked as designed, but they were never triggered on this path.
Google is launching a "universal" Gemini AI agent that works across apps and devices in the background. Announced at the Gemini at Work event, it will be available within the Gemini Enterprise app, where users can chat with the agent and assign it tasks from one interface.
Nudge Security, a vendor, promotes its product for governing OAuth grants that employees create when they approve third-party apps. The article cites an attack in which a compromised OAuth token from the third-party AI tool Context.ai let an attacker into Vercel after one employee had connected it to their Google Workspace account. It also reports that employees create an average of 88 OAuth grants each, 31 with data-level permissions.
Luca Guadagnino's film Artificial, a satirical biopic of OpenAI CEO Sam Altman, premiered at the New York Film Festival. Guadagnino said onstage that he is interested in people who want to play God. The film follows the factual events of Altman's rise to power at OpenAI and his brief ouster from the company.
The Verge's Decoder podcast interviews senior AI reporter Hayden Field about consumer AI agents, including Meta's Muse and OpenAI's Dots, which can handle tasks like restaurant reservations and inbox triage. The conversation covers the privacy and security implications of giving agents access to credit card information, email inboxes and other sensitive data.
Palo Alto Networks' Unit 42 and Zenity Labs researchers report that AWS has repeatedly patched autonomous agent security holes in AgentCore, and that the flaws have reappeared in slightly different forms. On September 18, 2026, Unit 42 reported that default configurations in AWS AgentCore Harness let attackers use prompt injection to steer the agent into exfiltrating plaintext credentials managed by AgentCore Identity, since the built-in shell tool, enabled by default, runs as root inside the harness. AWS closed that report as informative under its shared responsibility model.
Check Point's September 2026 threat report says organizations averaged 2,803 weekly cyber attacks, up 16% month over month and 48% year over year. Education was the most targeted sector at 6,656 weekly attacks per organization, and Europe had the sharpest regional rise at 61% year over year. One in every 39 enterprise GenAI prompts posed a high risk of sensitive data leakage, affecting 89% of organizations that regularly use GenAI tools.
Microsoft plans to add support for third-party deepfake detection solutions and a new impersonation protection feature to Teams meetings. According to Microsoft 365 Roadmap entries, both are in development and are slated for general availability in November after a worldwide rollout.
Pollo AI, a platform for turning creative ideas into images and videos, uses OpenAI models in its products, including its Pollo Agent workflow tool. The company reports more than 26 million users, and it routes tasks across GPT-5.6, GPT-6 Astra and GPT-Image-2.5 based on difficulty, reducing user time spent choosing or switching models by more than 50%.
Cybersecurity startup Rein Security announced a $25 million Series A round, bringing its total funding to $35 million. The round was co-led by Glilot Capital and Sienna Venture Capital, with support from Corner Ventures, Atlacle and RNP Capital Advisors. Rein, founded in 2024 and headquartered in Tel Aviv and New York City, extended its runtime protection platform to secure AI agents, which the company says already protect thousands of agents across multiple industries.
The essay, written with Seva Gunitsky and originally published in Foreign Affairs, examines how AI helps autocratic rulers monitor and replace human subordinates. It cites Margarita Simonyan's claim that some RT anchors are AI-generated and Russian Digital Development Minister Maksut Shadaev's April 2025 remark that AI "can definitely replace half of officials." It also describes China's "Zero Trust AI system," deployed across roughly 30 counties and cities to cross-reference more than 150 government databases.
AWS has released Strands Box, an open-source sandbox for AI agents, in developer preview under the Apache 2.0 license. It combines operating system-level isolation with policies written in Dogwood, an AWS-developed policy language, that can factor in an agent's prior activity across tools, and currently supports Macs with Apple silicon running macOS 15 or later. Dogwood does not evaluate files accessed directly through an agent harness's built-in tools, and the shell and Python interpreters run outside the sandbox as a trusted process.
Cisco Talos identified an APT spear-phishing campaign against individuals affiliated with Taiwan research organizations. The actor impersonated Taiwan European Union Centre, NCCU Institute of International Relations and Taiwan Research Institute, using legitimate public event themes. The campaign also used an AitM framework that impersonated Google authentication pages to intercept credentials and MFA challenges in real time.
Sam Altman, chief executive of OpenAI, told Politico's Decoded podcast and newsletter that the world should accept some bad things happening for the benefits of AI and people's agency to use it. Guardian US columnist Moustafa Bayoumi responds by arguing that AI is already implicated in violent deaths and nearly started world war three.
The Co-op has placed staff at Co-op Legal Services under AI surveillance. Automated listening technology records, analyses and gives every probate call a percentage score, according to a whistleblower who calls the system "oppressive and dystopian".
Refugee gig workers in Kakuma, Kenya, complete data tasks for tech companies without knowing the client, the purpose of the work, or whether they will be paid. One worker, Grace, used ChatGPT to research translations of Christian hymns and fed her findings to AOP Connect, a crowdsourced research platform owned by RWS. Refugees cannot legally work in Kenya, which leaves them with few options.
Fix: Nudge Security provides OAuth grant visibility, including dormant grants, and surfaces API keys, service accounts and remote MCP server connections. The source does not describe any other mitigation beyond this product.
BleepingComputer