Loading page…
Loading page…
Showing saved headlines while the latest news refreshes. Last successful update: .
News stories updated
Two U.S. House committees are urging the federal government to restrict Chinese-developed open-weight AI models from federal systems, federal contract work and high-risk critical infrastructure. The recommendations follow a seven-month investigation into model provenance, cybersecurity, unauthorized distillation and the growing use of Chinese models by American companies including Airbnb, Anysphere and DoorDash. The report proposes 12 policy actions, including an American open-weight AI initiative, an AI bill of materials, new provenance standards, stronger threat sharing and tighter federal controls. The proposals are not enacted law, but they signal a widening U.S.–China technology contest in which AI model supply chains are beginning to receive scrutiny once reserved for semiconductors, telecommunications equipment and critical software. Read More...
European Central Bank President Christine Lagarde has identified AI-amplified cyber threats as one of the most serious risks facing global finance, warning that increasingly capable autonomous systems could make attacks faster, cheaper and harder to attribute. Speaking at the European Banking Congress in Frankfurt on November 20, Lagarde said the combination of artificial intelligence and cyber threats is a top concern and argued that policymakers have not devoted enough attention to the issue. Her warning aligns with the ECB’s November Financial Stability Review, which found that agentic AI could increase the scale and sophistication of attacks while financial institutions remain exposed to concentrated cloud, software and third-party technology dependencies. Read More...
Dashlane says enterprise adoption of AI agents is creating a rapidly expanding credential-security problem. In a survey of 3,000 knowledge workers and 800 IT and security executives across the United States, United Kingdom and France, 72% of organizations reported at least one security incident involving an AI agent’s credentials or access during the previous year. Seventy-one percent said agents had standing access to internal systems or data, while 21% of workers admitted sharing passwords, API keys or login credentials with AI tools. The findings are self-reported rather than independently verified breach statistics, but they highlight a growing problem: enterprises are giving autonomous systems increasingly powerful access before identity, least-privilege and monitoring controls have fully caught up. Read More...
The U.S. Navy has selected SafeLogic for a paid post-quantum cryptography prototype pilot intended to inventory and prioritize vulnerable cryptographic assets across naval software environments. SafeLogic says its Cryptographic Posture Management platform will help NIWC Atlantic identify algorithms, certificates, protocols and dependencies that may need migration to NIST-approved quantum-resistant standards. The Navy’s broader procurement requirements independently confirm that automated cryptographic discovery is a central objective, but SafeLogic’s individual selection is currently company-reported. The pilot marks an important shift from discussing future quantum threats to mapping the actual cryptography embedded in long-lived government systems before migration begins. Read More...
Microsoft has launched an AI Transparency Initiative with more than 40 organizations, including Accenture, Adobe, KPMG, Mastercard and Volvo Cars, to develop common approaches for auditing autonomous AI models and agents. The initiative focuses on three questions: what an AI system was authorized to do, what it actually did and whether its actions can be independently verified. Microsoft argues that conventional logs often capture surrounding system activity without providing enough evidence to reconstruct an agent’s decision path or delegated authority. The collaboration could influence enterprise AI security, procurement, financial compliance and internal audit requirements, although it is not yet a finalized standard or certification regime. Read More...
OpenAI and Microsoft released major new AI systems on November 17 that could significantly expand the role of artificial intelligence in cybersecurity. OpenAI’s GPT-6.2 Astra and Sol both meet the company’s Critical cybersecurity capability threshold, while Astra is its first externally released model assessed at the High threshold for AI self-improvement. Microsoft simultaneously introduced MAI Cyber 2 Flash and expanded Microsoft Perception, a multi-agent security platform coordinating offensive testing, threat investigation and remediation. The releases highlight a rapidly changing security environment in which AI can discover vulnerabilities faster and operate continuously, but stronger capability also increases the importance of authorization, monitoring, human approval and cost control. Read More...
Thirteen Democratic senators on the Senate Commerce Committee are urging the Federal Aviation Administration to pause deployment of new Juniper networking equipment after a November 13 network disruption affected communications and radar feeds at 17 of the FAA’s 22 en route control centers. The lawmakers’ letter says 3,648 routers are planned for conversion, about 400 have been installed and only 69 are currently operational. The FAA says the precise cause of the Arkansas router instability remains under investigation and continues to report no evidence of an external cyberattack. The dispute is shifting attention from the outage itself to whether critical-infrastructure modernization programs have adequate testing, isolation, rollback and failover controls before new equipment enters production. Read More...
Cyber incidents reported by Marsh clients more than doubled between two recent two-week periods, rising from 48 in late May to 105 in late October 2026. Marsh says attackers are increasingly using AI-enabled and agentic tools to identify vulnerabilities, automate reconnaissance and exploit familiar weaknesses faster, while incident-response teams face growing capacity pressure. The data do not prove that AI caused every attack or that global cyber incidents rose at the same rate, but they illustrate an important shift: artificial intelligence may be compressing the time between vulnerability discovery and exploitation. That has major implications for cloud providers, banks, crypto infrastructure operators and companies deploying autonomous agents. Read More...
Hitachi has developed a prototype called Proof of Human Delegation, or PoHD, designed to prove that an AI agent was genuinely authorized by a specific person to perform defined actions. The system combines biometric authentication with a digitally signed delegation document using Hitachi’s Public Biometric Infrastructure, which reconstructs a signing key only when the user verifies their identity and deletes it after use. Hitachi demonstrated PoHD in a procurement workflow and plans to evaluate it for banking, payments, industrial operations, mobility and energy. The technology could address a growing security problem as autonomous agents gain permission to move money, manage systems and interact with critical infrastructure. Read More...
United Airlines grounded departing aircraft on November 16 amid a systemwide technology disruption, following a nationwide FAA ground stop, a major Microsoft 365 outage and widespread banking-service problems. The rapid succession of failures has raised questions about cybersecurity, artificial intelligence, foreign interference and the resilience of America's critical infrastructure. Federal authorities and affected companies have identified technical problems in several incidents, with no verified evidence of a coordinated cyberattack or Iranian involvement. Yet the disruptions expose a deeper concern: interconnected financial, transportation and cloud systems may be increasingly vulnerable to ordinary failures, insecure modernization and concentrated technology dependencies. Katharos examines what is known, what remains unanswered and why infrastructure security requires fundamental reform. Read More...
Israeli-American cybersecurity startup Corma has raised a $125 million Series A at a reported $1 billion valuation, only three months after emerging from stealth with a $60 million seed round. Led by New Enterprise Associates with participation from Sequoia Capital, Khosla Ventures and Coatue, the financing brings total disclosed funding to $185 million. Corma is building proprietary foundation models designed specifically for defensive cybersecurity and deploying them through autonomous agents that work across enterprise tools, logs and security telemetry. The funding reflects growing investor conviction that cybersecurity may require domain-specific AI rather than general-purpose models, but Corma’s performance claims remain company-reported rather than independently established benchmarks. Read More...
Federal Reserve Governor Michael Barr warned on November 16 that increasingly autonomous AI agents could create new financial-stability risks by accelerating deposit withdrawals, synchronizing trading behavior and acting inside customer accounts without sufficiently reliable authorization controls. Speaking in Milan, Barr said banks are already confronting situations in which AI systems attempt transactions or account actions that institutions cannot easily verify as genuinely authorized. He also warned that autonomous agents could amplify market manipulation, cyberattacks and liquidity stress. The remarks do not introduce a new regulation, but they sharpen the emerging policy debate around identity, permissioning and emergency controls for financial AI. Read More...
London-based quantum technology company Moth says its new Tessa encoding method stored and retrieved 1,568 numerical values across all 120 qubits of an IBM Quantum Nighthawk r2 processor, more than 12 times the 128 values recovered in a previously reported comparable experiment. Tessa uses the processor’s existing connectivity without adding SWAP operations or auxiliary qubits, and Moth has demonstrated the method on images, audio, DNA and protein sequences. The result addresses a real bottleneck in quantum machine learning: getting classical data onto noisy quantum hardware efficiently. The work is documented in an arXiv preprint but has not yet been independently reproduced or peer-reviewed. Read More...
South Africa will host the 10th International Conference on Quantum Techniques in Machine Learning from December 6–11 at Stellenbosch University, bringing together researchers from IBM, Google, ETH Zürich, MIT, Quantinuum and other leading institutions. QTML 2026 will feature more than 50 talks and about 150 research posters spanning quantum neural networks, hybrid quantum-classical algorithms, generative models, error correction and attempts to identify meaningful quantum advantage in machine learning. The event places South Africa inside a fast-growing global research field connecting quantum computing and artificial intelligence. It also highlights a crucial distinction: advances in quantum machine learning do not mean current encryption or blockchain security has been defeated. Read More...
Morocco will host the ECOWAS Women in Cybersecurity initiative in Rabat from November 16–20, bringing together female cybersecurity professionals, national incident-response teams and regional institutions for technical training, cyber crisis simulations and cross-border threat-intelligence exercises. The program is led by ECOWAS and Smart Africa with support from the World Bank and Morocco’s cybersecurity authorities. Its agenda includes critical-infrastructure protection, AI applications in cybersecurity, incident-response governance and regional coordination. The initiative is strategically important because African governments are expanding digital payments, cloud systems, AI infrastructure and public digital services at the same time that ransomware, supply-chain compromise and attacks on critical systems are becoming more sophisticated. Read More...