Governance
Why Liability and Insurance Won’t Save AI: Lessons From Cyber Insurance
(Daniel Schwarcz, Josephine Wolff – Lawfare – 9 September 2025) In 2024, California’s state legislature considered a bill, SB 1047, or the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, that would have, among other provisions, imposed liability on some AI companies for harm caused by their machine learning models. Ultimately, California Gov. Gavin Newsom vetoed the bill, but its proponents contend that imposing legal liability is the best way to force AI companies to make their machine learning models safer and more secure. That’s not an uncommon viewpoint these days—that the most effective path to promoting AI safety is to make sure that the people developing it are held responsible for any harms that their systems cause. But the history of data breaches, cybersecurity, and cyber insurance offers some cautionary lessons about just how useful liability is likely to be for making AI safer given the key role that liability insurers will inevitably play in pricing and covering this risk. – https://www.lawfaremedia.org/article/why-liability-and-insurance-won-t-save-ai–lessons-from-cyber-insurance
The AI Deregulation Agenda Has Helped Create an AI Bubble and May Hasten a Crash
(Amber Sinha – Tech Policy Press – 9 September 2025) A little over a month ago, Euractiv journalist Thomas Moller-Nielsen warned that the global trade war instigated by United States President Donald Trump takes attention away from a much graver economic threat — the Trump administration’s deregulation agenda. His analysis drew from a recent speech by US Federal Reserve Governor Michael Barr that pointed out that instances of significant deregulation preceded “all three of the most infamous financial meltdowns over the past century, namely the Great Depression in the 1930s, the ‘Savings & Loan’ crisis in the late 1980s and early 1990s, and the 2009 global financial crisis.” The same arguments may be applicable to digital deregulation, which may exacerbate risks to the market from unregulated artificial intelligence products and services. – https://www.techpolicy.press/the-ai-deregulation-agenda-has-helped-create-an-ai-bubble-and-may-hasten-a-crash/
AI adoption drops at large US companies for the first time since 2023
(DigWatch – 9 September 2025) Despite the hype surrounding AI, new data suggests corporate adoption of AI is slowing. A biweekly survey by the US Census Bureau found AI use among firms with over 250 employees dropped from nearly 14 percent in mid-June to under 12 percent in August, marking the largest decline since the survey began in November 2023. – https://dig.watch/updates/ai-adoption-drops-at-large-us-companies-for-the-first-time-since-2023 – https://futurism.com/ai-hype-automation-decline
AI threatens the future of entry level jobs
(DigWatch – 9 September 2025) The rise of AI puts traditional entry-level roles under pressure, raising concerns that career ladders may no longer function as they once did. Industry leaders, including Anthropic CEO Dario Amodei, warn that AI could replace half of all entry-level jobs as machines operate nonstop. A venture capital firm, SignalFire, found that hiring for graduates with under one year of experience at major tech firms fell by 50% between 2019 and 2024. The decline has been consistent across business functions, from sales and marketing to engineering and operations. – https://dig.watch/updates/ai-threatens-the-future-of-entry-level-jobs – https://www.cnbc.com/2025/09/07/ai-entry-level-jobs-hiring-careers.html
Geostrategies
National cyber director: U.S. strategy needs to shift cyber risk from Americans to its adversaries
(Tim Starks – Cyberscoop – 9 September 2025) The United States needs a “new, coordinated strategy” to counter its cyber adversaries and “shift the burden of risk in cyberspace from Americans to them,” National Cyber Director Sean Cairncross said Tuesday. “Collectively, we’ve made great progress in identifying, responding to and remediating threats, but we still lack strategic coherence and direction,” he said at the Billington Cybersecurity Summit. “A lot has been done, but it has not been sufficient. We’ve admired the problem for too long, and now it’s time to do something about it.” – https://cyberscoop.com/us-cybersecurity-strategy-sean-cairncross-shift-risk-china-trump-biden-cisa/
New project expands AI access for African languages
(DigWatch – 9 August 2025) Africa is working to close the AI language gap, as most global tools remain trained on English, Chinese, and European languages. The African Next Voices project has created the continent’s largest dataset of spoken African languages, covering 18 tongues across Kenya, Nigeria, and South Africa. Supported by a $2.2m Gates Foundation grant, the dataset includes 9,000 hours of speech in farming, health, and education settings. – https://dig.watch/updates/new-project-expands-ai-access-for-african-languages – https://www.bbc.com/news/articles/crkzgkkpx0lo
Security
Threat Actor Accidentally Exposes AI-Powered Operations
(Alessandro Mascellino – Infosecurity Magazine – 9 September 2025) A threat actor has unintentionally revealed their methods and day-to-day activities after installing Huntress security software on their own operating machine. The unusual incident gave analysts a remarkable inside look into how attackers use artificial intelligence (AI), research tools and automation to refine their workflows. – https://www.infosecurity-magazine.com/news/threat-actor-exposes-operations/
Salty2FA Phishing Kit Unveils New Level of Sophistication
(Alessandro Mascellino – Infosecurity Magazine – 9 September 2025) A phishing campaign leveraging the Salty2FA kit has been uncovered by cybersecurity researchers, revealing advanced techniques that highlight the growing professionalism of cybercrime operations. The kit demonstrates a high degree of technical innovation, with layered defenses designed to bypass traditional detection. – https://www.infosecurity-magazine.com/news/salty2fa-phishing-kit/
Axios User Agent Helps Automate Phishing on “Unprecedented Scale”
(Phil Muncaster – Infosecurity Magazine – 9 September 2025) Security experts have warned of a huge uptick in automated phishing activity abusing the Axios user agent and Microsoft’s Direct Send feature. ReliaQuest claimed in a new report today that it observed a 241% increase in phishing activity using Axios between June and August 2025. Axios accounted for nearly a quarter (24%) of all malicious user-agent activity analyzed in the period, making it 10 times more common than any other agents tracked by ReliaQuest. The threat intelligence vendor said Axios-powered attacks had a 58% success rate versus just 9% for incidents without the user agent. – https://www.infosecurity-magazine.com/news/axios-user-agent-automate-phishing/
LunaLock Ransomware threatens victims by feeding stolen data to AI models
(Pierluigi Paganini – Security Affairs – 9 September 2025) A new ransomware group, named LunaLock, appeared in the threat landscape with a unique cyber extortion technique, threatening to turn stolen art into AI training data. Recently, the LunaLock group targeted the website Artists&Clients and stole digital art. The group demanded $50K to the victims, threatening leaks and the use of the stolen data to train large language models (LLMs). “We have breached the website Artists&Clients to steal and encrypt all its data. If you are a user of this website, you are urged to contact the owners and insist that they pay our ransom. If the ransom is not paid, we will release all data publicly on this Tor site, including source code and personal data of users. Additionally, we will submit all artwork to AI companies to be added to training datasets.” reads the announcement published by the ransomware group on its Tor data leak site. – https://securityaffairs.com/182014/malware/lunalock-ransomware-threatens-victims-by-feeding-stolen-data-to-ai-models.html
Frontiers
Why accessibility might be AI’s biggest breakthrough
(Ben Edwards – Ars Technica – 9 September 2025) While tech companies market AI as a productivity tool for everyone, a UK government study reveals an unexpected result: Neurodiverse employees may be benefiting far more from chatbots than their neurotypical colleagues. The UK’s Department for Business and Trade recently released evaluation results from its Microsoft 365 Copilot trial showing that while overall satisfaction was 72 percent, neurodiverse employees reported statistically higher satisfaction (at a 90 percent confidence level) and were more likely to recommend the tool (at a 95 percent confidence level) than other respondents. “It’s leveled the playing field,” one participant with ADHD told researchers during follow-up interviews. One user with dyslexia said that the tool “empowered” them to perform tasks with confidence they previously lacked, particularly in report writing. Another dyslexic participant drew direct comparisons to existing accessibility software, noting that Copilot “does a hell of a lot more” than traditional assistive technology while being “embedded in your applications” rather than requiring separate programs. – https://arstechnica.com/information-technology/2025/09/study-finds-neurodiverse-workers-more-satisfied-with-ai-assistants/
Superconducting qubits power Stanford’s quantum router advance
(DigWatch – 9 September 2025) Quantum computers could become more efficient with a new quantum router that directs data more quickly within machines. Researchers at Stanford have built the component, which could eventually form the backbone of quantum random access memory (QRAM). The router utilises superconducting qubits, controlled by electromagnetic pulses, to transmit information to quantum addresses. Unlike classical routers, it can encode addresses in superposition, allowing data to be stored in two places simultaneously. – https://dig.watch/updates/superconducting-qubits-power-stanfords-quantum-router-advance
AI-tissue collaboration could transform drug trials and precision medicine
(DigWatch – 9 September 2025) Researchers combine human tissue models with explainable AI to analyse patient data and identify treatments that work best for specific patients. First applied to inflammatory bowel disease, the approach could improve clinical trial success rates and accelerate drug discovery. REPROCELL, IBM, and the STFC Hartree Centre have developed Pharmacology-AI, a platform uniting tissue models with machine learning. Delivered through the Hartree National Centre for Digital Innovation, it reduces costs, enhances trial design, and enables more targeted therapies. – https://dig.watch/updates/ai-tissue-collaboration-could-transform-drug-trials-and-precision-medicine – https://www.drugtargetreview.com/article/185362/ai-meets-human-tissue-to-fast-track-precision-medicine-development/
AI Mode in Google Search adds multilingual support to Hindi and four more languages
(DigWatch – 9 September 2025) Google has announced an expansion of AI Mode in Search to five new languages, including Hindi, Indonesian, Japanese, Korean and Brazilian Portuguese. The feature was first introduced in English in March and aims to compete with AI-powered search platforms such as ChatGPT Search and Perplexity AI. – https://dig.watch/updates/ai-mode-in-google-search-adds-multilingual-support-to-hindi-and-four-more-languages