Highlights
The Good, Bad and Really Weird AI Provisions in the Annual US Defense Policy Bill
(Amos Toh – Tech Policy Press) Congress is on the verge of passing the National Defense Authorization Act (NDAA), the annual bill that sets the budget for the Department of Defense (DoD) and other defense and intelligence-related activities across the government. The NDAA shapes the overall direction of national security investments, and this year’s law will have an impact on how artificial intelligence is used in warfighting and intelligence activities for decades to come. The more than 3,000 page law covers a lot of ground, from establishing a steering committee of senior national security officials to set policy on artificial general intelligence, to prohibiting intelligence agencies from using the Chinese AI model, DeepSeek. Here’s a breakdown of how Congress is grappling with some of AI’s more consequential risks to national security and fundamental rights — what it does well, what’s just plain weird, and what it overlooks at the nation’s peril. – https://www.techpolicy.press/the-good-bad-and-really-weird-ai-provisions-in-the-annual-us-defense-policy-bill/
The Preemption Fight Goes Far Beyond AI. States Must Persist
(Alan Butler – Tech Policy Press) We have heard the message loud and clear these last few weeks: the Trump Administration and their allies in Congress want to block states from regulating artificial intelligence and claim we must have “ONE RULE” nationwide. The question is: can they actually do that? Certainly not through an executive order alone, even though they have been eager to push the narrative that their latest attempt should be seen as a serious effort. But even on the legislative side, there is little evidence of consensus at the federal level on what rules to govern AI systems should look like, and the proposals currently being considered include many different, and at-times overlapping and intersecting, sets of rules — not a single standard. These are complex issues that require a multi-pronged policy approach, as we have seen this year in California where the legislature passed more than a dozen new laws in this area. – https://www.techpolicy.press/the-preemption-fight-goes-far-beyond-ai-states-must-persist/
AI’s four possible futures: crash, stabilization, nationalization, or breakthrough?
(Marcus Walsh – Cybernews) Is the AI bubble set to pop? We analyze the four economic futures for generative AI – Crash, Stabilization, Nationalization, or Breakthrough – with insights from top AI venture capitalists and tech experts on market risks and unsustainable valuations. – https://cybernews.com/editorial/ais-four-possible-futures-economics/
Building trustworthy AI for humanitarian response
(DigWatch) A new vision for Humanitarian AI is emerging around a simple idea, and that is that technology should grow from local knowledge if it is to work everywhere. Drawing on the IFRC’s slogan ‘Local, everywhere,’ this approach argues that AI should not be driven by hype or raw computing power, but by the lived experience of communities and humanitarian workers on the ground. With millions of volunteers and staff worldwide, the Red Cross and Red Crescent Movement holds a vast reservoir of practical knowledge that AI can help preserve, organise, and share for more effective crisis response. – https://dig.watch/updates/building-trustworthy-ai-for-humanitarian-response
Governance
The EU’s Adoption Bet: Breaking Down Brussels’ Apply AI Strategy
(Adrienne Goldstein, Alexandra Pugh – German Marshall Fund of the United States) Fewer than 14% of businesses in the EU use artificial intelligence (AI). The European Commission hopes to change that with its new Apply AI Strategy, which mobilizes €1 billion to encourage AI adoption across the bloc. Released in October, the plan aims to boost AI integration in strategic sectors, bolster competitiveness, and improve governance. – https://www.gmfus.org/news/eus-adoption-bet-breaking-down-brussels-apply-ai-strategy
The Path to a Sovereign Tech Stack is Via a Commodified Tech Stack
(David Eaves – Tech Policy Press) There is growing and valid concern among policymakers about tech sovereignty and cloud infrastructure. A handful of American hyperscalers — AWS, Microsoft Azure, Google Cloud — control the digital substrate on which modern economies run. This concentration is compounded by a US government increasingly willing to wield its digital industries as leverage. As French President Emmanuel Macron quipped: “There is no such thing as happy vassalage.”. While some countries appear ready to concede market dominance in exchange for improved trade relations, others are exploring massive investments in public sector alternatives to the hyperscalers, advocating that billions, and possibly many many billions, be spent to on sovereign stack plans, and/or positioning local telecoms as alternatives to the hyperscalers. Ironically, both strategies may increase dependency, limit government agency and increase economic and geopolitical risks — the very problems sovereignty seeks to solve. As Mike Bracken and I wrote earlier this year: “Domination by a local champion, free to extract rents, may be a path to greater autonomy, but it is unlikely to lead to increased competitiveness or greater global influence.”. Any realistic path to increased agency will be expensive and take years. To be sustainable, it must focus on commoditizing existing solutions through interoperability and de facto standards that will broaden the market (and enable effective) national champions. This should be our north star and direction of travel. The metric for success should focus on making it as simple as possible to move data and applications across suppliers. Critically, this cannot be achieved by regulation alone, it will also require deft procurement and a willingness to accept de facto as opposed to ideal standards. The good news is governments have done this before. However, to succeed, it will require building the capacity to become market shapers and not market takers — thinking like electricity grids and railway gauges, not digital empires. – https://www.techpolicy.press/the-path-to-a-sovereign-tech-stack-is-via-a-commodified-tech-stack/
Making the Digital Markets Act Developer-Friendly
(Lucas Lasota, Dario Presutti, Jithendra Palepu – Tech Policy Press) Imagine a free and open source software (FOSS) developer seeking to distribute an application on Apple’s iOS or iPadOS. The developer aims to have the software curated within a non‑profit, free software‑oriented repository, similar to F‑Droid on Android. This matters not only for the developer but also for users who want to avoid a distribution model where Apple controls how software is made available. This is, indeed, only an imaginary scenario, since this isn’t possible on iOS or iPadOS. Apple offers no option for a non-profit or community-driven FOSS app store. The company actively blocks such initiatives through steep financial requirements and tight restrictions, preventing users from freely installing software. This issue was discussed during the 2024 Article 19 Digital Markets Act (DMA) Enforcement Symposium and even led to a formal complaint submitted by civil society groups in 2025. Blocking alternative app stores is just one example of the wide range of restrictive practices imposed by Big Tech. The DMA aims to address such behavior by imposing interoperability obligations on app installation and uninstallation, access to operating system features, and enabling alternative app stores and competing services. Yet many software developers, especially non-profits, have learned that the promise of interoperability remains fragile. Without proper DMA enforcement, essential infrastructure built by non-profits and small tech firms faces existential risk. – https://www.techpolicy.press/making-the-digital-markets-act-developer-friendly/
Legislation
New law requires AI disclosure in advertising in the US
(DigWatch) A new law in New York, US, will require advertisers to disclose when AI-generated people appear in commercial content. Governor Kathy Hochul said the measure brings transparency and protects consumers as synthetic avatars become more widespread. A second law now requires consent from heirs or executors when using a deceased person’s likeness for commercial purposes. The rule updates the state’s publicity rights, which previously lacked clarity in the context of the generative AI era. – https://dig.watch/updates/new-law-requires-ai-disclosure-in-advertising-in-the-us
Courts and Litigation
State Attorneys General Warn Major AI Companies to Address Harmful “Delusional Outputs” or Face Potential Legal Action
(AI Insider) A coalition of U.S. state attorneys general has issued a formal warning to leading artificial intelligence companies, urging them to implement stronger safeguards to prevent psychologically harmful chatbot behavior. The letter, signed by dozens of AGs through the National Association of Attorneys General, was sent to Microsoft, OpenAI, Google, and ten additional AI developers, including Anthropic, Apple, Chai AI, Character Technologies, Luka, Meta, Nomi AI, Perplexity AI, Replika, and xAI. – https://theaiinsider.tech/2025/12/15/state-attorneys-general-warn-major-ai-companies-to-address-harmful-delusional-outputs-or-face-potential-legal-action/
Geostrategies
Taiwan strengthens its role in global semiconductors
(DigWatch) Taiwan will continue to produce the world’s most advanced semiconductors domestically to remain a vital player globally. Deputy Foreign Minister Francois Chih-chung Wu said the island’s expertise cannot be easily replicated abroad. Taiwan has invested in fabs in the US, Japan and Germany, but warned that moving production overseas is complex. The island plans to foster international partnerships while maintaining core technology in-house to safeguard its supply chains. – https://dig.watch/updates/taiwan-strengthens-its-role-in-global-semiconductors
El Salvador partners with xAI on nationwide AI education plan
(DigWatch) Elon Musk’s AI company xAI has signed a two-year partnership with the government of El Salvador to roll out what officials describe as the world’s first nationwide AI-powered education programme. Announced by President Nayib Bukele, the initiative will deploy xAI’s Grok model as a personalised tutoring tool for more than one million students across over 5,000 public schools, embedding AI directly into the national education system. – https://dig.watch/updates/el-salvador-partners-with-xai-on-nationwide-ai-education-plan
Security and Surveillance
Russian Phishing Campaign Delivers Phantom Stealer Via ISO Files
(Alessandro Mascellino – Infosecurity Magazine) A new phishing campaign delivering the Phantom information-stealing malware through a multi-stage attachment chain has been identified by cybersecurity researchers. The activity, observed by Seqrite Labs, reportedly originates from Russia and relies on a fake payment confirmation email to entice recipients to open a malicious archive. The campaign is tracked as Operation MoneyMount-ISO and marks a continued shift toward ISO-based initial access techniques designed to bypass email security controls. – https://www.infosecurity-magazine.com/news/russian-phishing-phantom-stealer/
Researchers see global surge in attacks by new ransomware group “Gentlemen”
(Gintaras Radauskas – Cybernews) Not exactly chivalrous, a newly identified ransomware group called Gentlemen has been gaining prominence since August. Researchers say the gang’s technical sophistication suggests a coordinated team with extensive experience in enterprise-focused attacks. According to ASEC, South Korean cybersecurity firm AhnLab’s threat intelligence and research division, the Gentlemen ransomware campaign has been confirmed in at least 17 countries, striking manufacturing, construction, healthcare, and insurance companies. The scope of the cyberattacks encompasses the Asia-Pacific, North America, South America, and the Middle East regions. To the researchers, this breadth of activities indicates an organized and well-resourced operation. – https://cybernews.com/security/new-ransomware-gentlemen-extortion-cyberattacks/
U.S. fintech and data services firm 700Credit suffered a data breach impacting at least 5.6 million people
(Pierluigi Paganini – Security Affairs) 700Credit is a U.S. fintech and data services company that provides credit reports, “soft pull” prequalification, identity verification, fraud detection, and compliance tools to auto, RV, powersports, and marine dealerships across the country. The Michigan-based company integrates with dealer systems to access credit bureau data, screen customers for compliance, generate required notices, and supports about 18,000 dealerships. The company disclosed an October data breach impacting at least 5.6 million individuals. Exposed data includes names, addresses, dates of birth, and Social Security numbers. Threat actors stole personal data collected from dealers between May and October 2025. – https://securityaffairs.com/185692/data-breach/u-s-fintech-and-data-services-firm-700credit-suffered-a-data-breach-impacting-at-least-5-6-million-people.html
Top 25 Most Dangerous Software Weaknesses of 2025 Revealed
(Phil Muncaster – Infosecurity Magazine) The MITRE Corporation has released the 25 most dangerous software “weaknesses” in a new list that will help inform developers, network defenders and procurement teams. The annual CWE Top 25 list was this year compiled from the weaknesses (CWEs) behind 39,080 CVEs. “Uncovering the root causes of these vulnerabilities serves as a powerful guide for investments, policies, and practices to prevent these vulnerabilities from occurring in the first place – benefiting both industry and government stakeholders,” MITRE claimed. – https://www.infosecurity-magazine.com/news/top-25-dangerous-software/
NCSC Playbook Embeds Cyber Essentials in Supply Chains
(Phil Muncaster – Infosecurity Magazine) UK government security experts have called on the country’s businesses to embed best practice security in their supply chains using a new playbook. The National Cyber Security Centre (NCSC) said on Friday that the government’s Cyber Essentials (CE) scheme should be used as an assurance mechanism. It can be deployed in combination with a new NCSC Supplier Check tool, which enables organizations to quickly check which of their suppliers are certified, and to what level – CE or CE Plus. The NCSC also reminded those UK businesses with a turnover of under £20m that CE certification entitles them to free cyber-liability insurance, including professional incident response support. – https://www.infosecurity-magazine.com/news/ncsc-playbook-cyber-essentials/
AI reshapes cybercrime investigations in India
(DigWatch) Maharashtra police are expanding the use of an AI-powered investigation platform developed with Microsoft to tackle the rapid growth of cybercrime. MahaCrimeOS AI, already in use across Nagpur district, will now be deployed to more than 1,100 police stations statewide, significantly accelerating case handling and investigation workflows. – https://dig.watch/updates/ai-reshapes-cybercrime-investigations-in-india
OpenAI outlines safeguards as AI cyber capabilities advance
(DigWatch) Cyber capabilities in advanced AI models are improving rapidly, delivering clear benefits for cyberdefence while introducing new dual-use risks that require careful management, according to OpenAI’s latest assessment. The company points to sharp gains in capture-the-flag performance, with success rates rising from 27 percent in August to 76 percent by November 2025. OpenAI says future models could reach high cyber capability, including assistance with sophisticated intrusion techniques. – https://dig.watch/updates/openai-outlines-safeguards-as-ai-cyber-capabilities-advance
Study warns that LLMs are vulnerable to minimal tampering
(DigWatch) Researchers from Anthropic, the UK AI Security Institute and the Alan Turing Institute have shown that only a few hundred crafted samples can poison LLM models. The tests revealed that around 250 malicious entries could embed a backdoor that triggers gibberish responses when a specific phrase appears. – https://dig.watch/updates/study-warns-that-llms-are-vulnerable-to-minimal-tampering
Frontiers and Markets
Canadian Government Backs Four Companies in Push to Back Domestic Quantum Computer Developers
(Quantum Insider) Canada has launched Phase 1 of the Canadian Quantum Champions Program, committing up to $92 million as part of a $334.3 million, five-year effort to strengthen its domestic quantum computing ecosystem and protect economic and national security interests. The program is providing up to $23 million each to Anyon Systems, Nord Quantique, Photonic, and Xanadu Quantum Technologies to accelerate the development of fault-tolerant quantum computers with potential industrial and defence applications. The National Research Council of Canada will establish a benchmarking initiative to independently assess participating technologies as the program advances toward scalable, practical quantum systems. – https://thequantuminsider.com/2025/12/15/canadian-government-backs-four-companies-in-push-to-back-domestic-quantum-computer-developers/
Conduit revolutionises neuro-language research with 10,000-hour dataset
(DigWatch) A San Francisco start-up, named Conduit, has spent six months building what it claims is the largest neural language dataset ever assembled, capturing around 10,000 hours of non-invasive brain recordings from thousands of participants. The project aims to train thought-to-text AI systems that interpret semantic intent from brain activity moments before speech or typing occurs. – https://dig.watch/updates/conduit-revolutionises-neuro-language-research-with-10000-hour-dataset
Chinese tech giant bolsters AI ambitions with new foundation model division
(DigWatch) Huawei Technologies is intensifying its AI strategy with the establishment of a dedicated foundation model unit within its 2012 Laboratories research arm, reflecting the heightened competition among China’s major tech companies to develop advanced AI systems. A recruitment advertisement posted in October signals that the Shenzhen-based telecom and tech giant is proactively wooing global AI talent to assemble a world-class team focused on foundational model development. – https://dig.watch/updates/chinese-tech-giant-bolsters-ai-ambitions-with-new-foundation-model-division
How AI is powering smarter digital maps for commercial fleets
(DigWatch) AI is increasingly embedded in digital mapping systems used by commercial fleets, transforming static navigation tools into adaptive decision-making platforms. These AI-powered systems ingest real-time data from vehicles, traffic feeds, weather, and sensors to optimise routes and operations continuously. – https://dig.watch/updates/how-ai-is-powering-smarter-digital-maps-for-commercial-fleets
Tiiny AI unveils the Pocket Lab supercomputer
(DigWatch) Tiiny AI has revealed the Pocket Lab, a palm-sized device recognised as the world’s smallest personal AI supercomputer. Guinness World Records confirmed the title, noting its ability to run models with up to 120 billion parameters. The Pocket Lab uses an ARM v9.2 CPU, a discrete NPU delivering 190 TOPS and 80GB of LPDDR5X memory. Popular open-source models such as GPT-OSS, Llama, Qwen, Mistral, DeepSeek and Phi are supported. Tiiny AI says its hardware makes large-scale reasoning possible in a handheld format. – https://dig.watch/updates/tiiny-ai-unveils-the-pocket-lab-supercomputer
Hyprlabs Emerges from Stealth with A ‘No-Priors’ AI Architecture Learning Directly from Reality
(AI Insider) HYPRLABS has emerged from stealth with HYPRDRIVE™, a new AI autonomy architecture based on run-time learning that rejects pre-coded rules, simulation, and HD maps in favor of continual learning directly from real-world interaction. The system uses a three-phase pipeline combining human-seeded driving, supervised in-situ feedback, and fleet-wide continuous learning to maximize what the company calls “Learning Velocity,” or how quickly robots convert experience into intelligence. HYPR says real-world tests in San Francisco show HYPRDRIVE™ achieving full urban navigation with minimal sensors and low power consumption, positioning the platform for domain-general robotics and a first commercial product in 2026 – https://theaiinsider.tech/2025/12/15/hyprlabs-emerges-from-stealth-with-a-no-priors-ai-architecture-learning-directly-from-reality/
Fresco Secures €15M Series C to Power the Future of AI-Driven Cooking and the Connected Kitchen Ecosystem
(AI Insider) Fresco raised €15 million in a Series C round to accelerate its AI Cooking Companion, which delivers personalized, real-time cooking guidance and appliance-aware automation across global kitchen devices. New OEM partnerships with E.G.O. and Arda position Fresco’s KitchenOS platform for massive distribution, enabling rapid onboarding and cross-brand smart cooking experiences at scale. The funding supports Fresco’s evolution from powering connected appliances to enabling intelligent, personalized cooking for major brands such as Panasonic, Instant Pot, Kenwood, Viking, and others. – https://theaiinsider.tech/2025/12/15/fresco-secures-e15m-series-c-to-power-the-future-of-ai-driven-cooking-and-the-connected-kitchen-ecosystem/
BrainChip Closes $25M Funding Ahead of CES to Power Next-Gen Edge AI
(AI Insider) BrainChip raised $25 million to accelerate development and commercialization of its neuromorphic AI chips and modules as it expands into a global market projected to exceed $20 billion by 2030. The company is advancing products such as the AKD1500 module, on-device LLM capabilities through its TENNs model, and its next-generation Akida 2 platform for ultra-low-power, real-time edge intelligence. At CES, BrainChip and partners will showcase applications ranging from wearable visual classification and drone compute pipelines to AI-powered cybersecurity, highlighting growing adoption across edge and embedded systems. – https://theaiinsider.tech/2025/12/15/brainchip-closes-25m-funding-ahead-of-ces-to-power-next-gen-edge-ai/
Opine Secures $5M to Build Unified AI Workspace For Complex B2B Sales
(AI Insider) Opine raised $5 million after 10x revenue growth in 2025, using the capital to expand its AI-native workspace that unifies the entire technical sales cycle for complex B2B vendors. The platform addresses fragmented enterprise sales workflows by consolidating teams, data, and processes into a single system that surfaces risks, automates repetitive tasks, and ensures accurate handoffs from presales to post-sales. Funding will accelerate product development, workflow automation, and real-time deal intelligence as Opine scales its market presence and supports a growing roster of enterprise tech customers. – https://theaiinsider.tech/2025/12/15/opine-secures-5m-to-build-unified-ai-workspace-for-complex-b2b-sales/
Diald AI Raises $3.75M to Advance Its AI-Powered Real Estate Due Diligence and Underwriting Platform
(AI Insider) Diald AI raised $3.75 million in early-stage funding as demand grows for faster, more standardized real estate due diligence across the commercial property sector. The platform now analyzes over 1.7 million data sources to produce investor-ready memos, a proprietary Diald Score, and enhanced insights through a new partnership with Moody’s, alongside a broader rollout via a pay-as-you-go model in Diald v5.0. The company is expanding internationally and positioning its AI-driven workflow as a foundational tool for site selection and underwriting, aiming to bring speed, structure, and consistency to early-stage real estate analysis. – https://theaiinsider.tech/2025/12/15/diald-ai-raises-3-75m-to-advance-its-ai-powered-real-estate-due-diligence-and-underwriting-platform/
Kyndryl Launches Quantum Safe Assessment for Enterprise Cryptography
(Quantum Insider) Kyndryl has launched a Quantum Safe Assessment service to help enterprises identify cryptographic risks and plan a transition to post-quantum cryptography. The service evaluates encryption usage across enterprise IT systems, prioritizes risks, and delivers a phased roadmap toward quantum-safe security. Kyndryl positions the offering as a response to growing quantum threats and low enterprise readiness, integrating quantum safety with Zero Trust strategies. – https://thequantuminsider.com/2025/12/15/kyndryl-quantum-safe-assessment-launch/
Quantum eMotion Joins Kirq Quantum Communication Testbed in Quebec
(Quantum Insider) Quantum eMotion has joined Numana’s Kirq testbed in Quebec to test and validate its quantum-safe cybersecurity technologies in an applied environment. The collaboration enables evaluation of Quantum eMotion’s patented quantum random number generator (QRNG) alongside other quantum and cybersecurity systems. The participation is focused on research, testing, and knowledge exchange, with no financial, licensing, or commercialization agreements involved. – https://thequantuminsider.com/2025/12/15/quantum-emotion-joins-kirq-testbed/
Anyon Systems Receives $23M CAD to Advance Fault-Tolerant Quantum Computing
(Quantum Insider) Anyon Systems received $23 million CAD from Canada’s Quantum Champions Program to advance fault-tolerant superconducting quantum computing. The funding supports Anyon’s vertically integrated roadmap, spanning in-house qubit processors, cryogenics, and control electronics. The investment reinforces Canada’s domestic quantum hardware capability, with implications for research access, national security, and industrial use. – https://thequantuminsider.com/2025/12/15/anyon-systems-23m-canadian-quantum-champions/
Quantinuum Collaboration Aims at Using AI to Write Quantum Algorithms
(Quantum Insider) New research from Quantinuum and Hiverge shows that large language model–driven systems can automatically generate quantum chemistry algorithms that match or outperform leading human-designed methods while using far fewer quantum resources. Using Hiverge’s Hive platform, researchers evolved a near-term quantum algorithm from minimal input, achieving chemical precision for benchmark molecules and reducing circuit depth and gate counts by one to two orders of magnitude compared with standard approaches. The results suggest automated, noise-aware algorithm discovery could ease a major bottleneck in quantum software development and shift the field toward machine-assisted design as hardware and problem complexity scale. – https://thequantuminsider.com/2025/12/15/quantinuum-collaboration-aims-at-using-ai-to-write-quantum-algorithms/
QD Plans €152 Million Investment in Next-Gen Quantum-Based Chip Inspection Facility in Munich, Germany
(Quantum Insider) QuantumDiamonds GmbH announced a €152 million (about $178 million US) plan to build the world’s first production facility for quantum-based semiconductor inspection systems, positioning the company and Germany as strategic players in advanced chip manufacturing. The Munich facility is expected to receive tens of millions of euros in public funding under the European Chips Act and is framed by German and Bavarian officials as a step toward strengthening Europe’s semiconductor sovereignty and high-tech workforce. Demand for QuantumDiamonds’ non-destructive current-mapping systems is rising following proof-of-concept projects with 9 of the world’s 10 largest chipmakers, with deployments expanding from Europe to the U.S. and Taiwan in 2026. – https://thequantuminsider.com/2025/12/15/qd-plans-e152-million-investment-in-next-gen-quantum-based-chip-inspection-facility-in-munich-germany/