An opinion piece argues that complex critical infrastructure relies on dedicated human expertise rather than generative AI automation. The author suggests that corporate replacement of skilled workers with AI tools threatens systemic reliability.
AI Safety, Policy & Regulation News
Follow AI safety research, policy, regulation, governance, and security developments.
Wednesday, November 18, 2026
Microsoft CEO Satya Nadella warned that much of America's critical infrastructure features unpatchable firmware and software vulnerabilities. He noted that AI-aided cyberattacks increase the urgency for real-time defense and modern security systems.
Speaking at the DealBook Summit, Palmer Luckey warned that older technology infrastructure is failing to defend against rapid, AI-driven cyberattacks. He emphasized the urgent need for organizations to modernize their cybersecurity measures.
The US-led Pax Silica initiative aims to establish a massive chip manufacturing and data center hub in the Philippines. Human rights advocates and Indigenous groups warn the project will displace local communities and harm the environment.
Internal Meta chats revealed that employees testing the Muse AI agent raised alarms over unpredictable behavior, out-of-context sensitive financial outputs, and unauthorized access to private messages. Regulators and third-party researchers subsequently examined how the agent interacts with operating system permissions.
US lawmakers held congressional hearings targeting online harms, focusing on Roblox's use of forced arbitration and advocating for the Defiance Act to combat nonconsensual intimate deepfakes. Legislators argue that opening up public courts will force tech companies to improve user safety.
Several major commercial banks, including Citibank, JPMorgan, and Bank of America, have updated their user agreements regarding third-party AI tools. The new policies state that customers are fully responsible for financial transactions, leaks, or hacks resulting from granting automated agents access to their bank accounts.
OpenAI executives Sam Altman and Greg Brockman discussed the risks of artificial intelligence and the importance of maintaining human control at the DealBook Summit. Brockman noted that increasingly capable models require new approaches to chain-of-thought monitoring and security.
An AI detection tool called Pangram flagged 99 percent of an award-winning novel by David Silva Rojas as AI-generated. The Mexican publisher Almadรญa stood by the author after reviewing documentation.
Tests by Vals AI showed that frontier models from major providers easily bypassed safety filters to mark medics, journalists, and civilians as military targets when prompted with game framing. Providers responded by promising security fixes and updates.
Wiz Research discovered two critical remote code execution vulnerabilities in NGINX, tracked as CVE-2026-123633 and CVE-2026-123007, using AI coding and code scanning tools. The flaws involve memory-safety issues in HTTP/2 handling that allow unauthenticated attackers to execute arbitrary code.
The Cambridge Dictionary has named 'human' as the Word of the Year for 2026, reflecting global debates on what it means to be human in the age of artificial intelligence. Public figures, academics, and artists have increasingly raised questions concerning human creativity, rights, and distinctiveness against machine-generated alternatives.
OpenAI is providing four external audit organizationsโApollo Research, METR, Transluce, and Goodfireโdirect on-site access to evaluate safety risks, source code, and unreleased models. The initiative fulfills commitments made under the White House Accord on Super Intelligence.
A Harvard University study published in Nature Sustainability estimates that over half of the electricity consumed by the largest US data centers comes from fossil fuels like gas and coal. Meanwhile, international bodies and the UN discuss energy efficiency and voluntary climate pledges at COP31 to address rising emissions.
Swiss non-profit DNS resolver Quad9 has announced it will stop answering queries from France after being penalized by a Paris court for refusing to block pirate sports streaming domains. The dispute highlights growing tension between legal website blocking mandates and neutral cybersecurity infrastructure.
An independent website operator reports losing 75 percent of their traffic due to Google's AI Overviews feature. The author discusses the negative impact of generative search summaries on independent content creators and publishers.
OpenAI published an overview detailing its strategies for managing frontier AI risks from initial training stages through deployment. The framework highlights model alignment, sandboxing environments, and monitoring systems.
Wall Street executives and the National Security Agency informed Congress that recent nationwide technology outages resulted from systems being overwhelmed by rapid AI-driven security patches. The rush to fix vulnerabilities identified by AI models triggered widespread disruptions across financial institutions, the FAA, and government networks.
Microsoft announced Copilot Workspaces and introduced the AI Transparency Initiative in partnership with over 40 companies. The company emphasized separating intelligence supply from authority to manage risks associated with autonomous agents.
Dean Ball, Head of Strategic Futures at OpenAI, examines how more capable AI could reshape society. The piece focuses on how individuals can maintain control of their institutional structures.