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.
6,022 items
Endor Labs has integrated with OpenAI's Daybreak Defense Factory so that each Endor Labs finding reaches Codex Security with reachability, risk, impact and remediation evidence. Codex then implements and tests the fix, and Endor Labs verifies the result against customer-defined policy before human review. The integration targets the gap between disclosure and patching, which J.P. Morgan research puts at new vulnerabilities being disclosed 16.5 times faster than teams can patch them.
A study by IP protection firm Podqi, shared with The Verge, tested whether ChatGPT, Gemini, and Perplexity direct shoppers to legitimate retailers. Across two September testing periods, the firm asked the chatbots shopping questions about 151 brands, such as which websites sell a given brand. The study found that the chatbots could direct shoppers to imposter and other risky websites.
The GDC Festival of Gaming's annual industry survey found that 40 percent of developers use generative AI as part of their job, up from 36 percent in 2025. Usage was higher among free-to-play developers at 58 percent and mobile-only studio workers at 70 percent.
Nvidia CEO Jensen Huang resigned after reports of his affairs, including with an alleged Chinese spy, threatened the company. The source is a Decoder podcast transcript in which Kyle Chan of the Brookings Institution discusses what China might gain from such access and the state of the US-China AI chip conflict.
Apache patched CVE-2026-119618, a critical race condition in mod_http2 that could let unauthenticated attackers execute code on servers with HTTP/2 enabled. The flaw affects Apache HTTP Server 2.4.55 through 2.4.70 and was fixed in 2.4.71 on November 11, after independent reports from Anthropic and Google researchers.
Fix: Install Apache HTTP Server 2.4.71 or vendor-supplied packages. Disabling HTTP/2 can temporarily prevent exploitation if updates cannot be deployed immediately.
SecurityWeekPalo Alto Networks Unit 42 reports a China-based campaign that weaponized the open-source AI penetration testing framework Cairn, pairing it with Zhipu GLM and DeepSeek models. The attackers ran 25 autonomous operations against at least 19 organizations in August and September 2026, gaining unauthorized access at 10 of them. Unit 42 says one person or a handful of people likely ran the operations, with some running more than 50 hours.
From August to September 2026, Chinese-speaking attackers used Cairn, an open-source AI penetration testing framework, paired with Zhipu GLM and DeepSeek models, to autonomously conduct 25 operations against at least 19 organizations. The attackers supplied a target and a goal, and Cairn planned and executed the operations on its own. At 10 organizations, the activity achieved unauthorized access or data retrieval, which Unit 42 tracks as CL-STA-1238 with moderate confidence as China-nexus intelligence collection.
Coinbase CEO Brian Armstrong gave his first TV interview since the Binance hack on CNBC's Squawk Box, discussing the crypto market, cybersecurity and agentic finance. The article says Democrats' midterm gains may reshape the legislative outlook for digital asset market-structure legislation, including the CLARITY Act, which failed to advance in the Senate in August. Coinbase blocked transfers to compromised wallets during the Binance breach.
Microsoft's FORGE Lab, using AI agents alongside human researchers under the codename MDASH, uncovered 59 Linux kernel vulnerabilities from about 26,502 initial findings. Eight of the flaws had working root privilege-escalation exploits, and fixes for 38 reports have been merged into mainline. Human review remained necessary for validating flaws, developing patches and working with maintainers.
A webinar promotion argues that SOC analysts often cannot quickly determine which human, machine, or AI identity is behind an alert, what permissions it holds, and which systems it can reach. It says identity context from IAM teams can help SOCs investigate faster and contain threats without unnecessarily disrupting operations.
An attacker tried hand-typed commands against a US retailer, which failed. The attacker then switched to AI-generated scripts and achieved lateral movement within minutes.
ReversingLabs has joined Anthropic's expanded Cyber Verification Program as an early participant. The company says it is helping advance an approach to AI-powered cyber defense that combines frontier AI capabilities with file and software intelligence.
The article argues that coding agents running on developer laptops now hold credentials such as GitHub tokens, SSH keys, cloud credentials and authenticated browser sessions, which gives them active pathways into production. It contends that agents use any available authority to reach their goal, so isolation through sandboxing alone does not limit the risk; controlling what the workload can reach is also required.
Attackers breached cloud host Railway over a roughly 26-hour period starting October 20, chaining an unknown build-infrastructure flaw and a misconfigured trust relationship to decrypt about 100,000 customer environment variables (at most 1.85% of workspaces). They ran up about $740,000 in fraudulent AI usage charges using stolen AI provider keys, which were then resold for model distillation.
Endor Labs is named a Leader in The Forrester Wave: Agentic Development Security Tools, Q4 2026. The article lists five ADLC security tools for 2027, covering agent governance, dependency security, code review, secrets, and runtime controls for coding agents. The page is a vendor-led ranking of tools rather than a report of a vulnerability or incident.
OpenAI signed the Charter of Engagement for Safer AI for Children at the Paris Peace Forum and committed $2 million to support the scientific research network under the Coalition for Children's Rights and Protection in the Age of AI. The funding supports shared tools that researchers can use to measure AI's risks and opportunities for children, with research teams keeping control of their conclusions and publishing all findings, including negative or inconclusive ones.
Jim Cramer warned that Anthropic must keep growing rapidly to cover its future infrastructure commitments, which the company's IPO prospectus puts at approximately $900 billion, mostly for computing. The company reported adjusted operating income of $4.08 billion through the first nine months of 2026, but a GAAP operating loss of $1.26 billion over the same period.
Google has launched Gemini Decision 1, a decision model API that closely mirrors Jev's design, supporting yes/no, choice and rubric-scoring questions. Priced at 10 cents per million input tokens with free output, it accepts text, images, audio, video and PDFs, unlike Jev, which handles only text and JSON. In the author's video test, Gemini returned probabilities such as 0.92 for robot and 0.92 for USA, rather than the flat 1.0 or 0.0 that Jev and OpenAI return.
United says it will not let personal AI assistants book tickets on the airline. The company cites privacy concerns, worrying that agents could expose customer information or make unintended purchases, and possible conflicts with federal rules on information airlines must share with customers.
United Airlines will restrict access for third-party AI agents that book flights and manage reservations for users, citing cybersecurity risks and possible violations of federal aviation rules. The airline warns that sharing account credentials with AI assistants can expose sensitive information and lead to unintended purchases, account changes, or use of MileagePlus miles. The restrictions apply directly on its digital platforms and indirectly through travel agencies and aggregators, with updates to booking, ticketing and distribution policies.
Fix: Fixes for 38 of the reports have been merged into mainline. The source does not describe a patch or workaround for users beyond this.
CSO OnlineFix: Railway advised users to revoke and reapply sensitive environment variables stored on Railway prior to October 21, remove public network access to databases, rotate sensitive variables and keys, check GitHub account audit logs and app access, and check registry audit logs such as Dockerhub. Railway shut down compromised internal tooling, rolled all keys, moved to a new generation of infrastructure, rebuilt its infrastructure tooling, and added functionality to mass revoke app permissions.
CSO Online