Proposed US law would make AI risk reporting a legal obligation

Anthropic accuses Alibaba of running largest distillation campaign against Claude

Anthropic cáo buộc phòng thí nghiệm Qwen của Alibaba đã thực hiện chiến dịch "distillation" (tinh chỉnh mô hình) quy mô lớn nhất từ trước đến nay chống lại Claude, sử dụng gần 25.000 tài khoản giả mạo để tương tác 29 triệu lần trong giai đoạn 4-6/2026. Họ kêu gọi chính phủ Mỹ hành động, bao gồm kiểm soát xuất khẩu chip AI và trừng phạt hành vi distillation, trong bối cảnh chuẩn bị IPO và đang đối mặt lệnh hạn chế xuất khẩu từ chính phủ Mỹ.

Những thông tin về chiến dịch distillation quy mô lớn của Alibaba nhằm xâm phạm khả năng kỹ thuật của Claude không chỉ là cảnh báo về rủi ro cạnh tranh công nghệ mà còn là cảnh báo về những thách thức pháp lý, chính trị và chiến lược mới đối với các công ty AI, đặc biệt khi họ đang chuẩn bị IPO và đối mặt với các hạn chế xuất khẩu từ chính phủ.

Actress Cate Blanchett launches a free registry tool to keep AI from using your likeness

Nữ diễn viên Cate Blanchett và nghị sĩ Eva Maydell đã ra mắt Human Consent Registry tại Nghị viện Châu Âu, một công cụ miễn phí giúp người dùng đặt điều kiện về cách AI sử dụng hình ảnh, giọng nói và tên tuổi của họ. Mặc dù không có tính pháp lý bắt buộc, registry này cung cấp điểm tham chiếu minh bạch cho sự đồng thuận, mở rộng cho cá nhân, đại lý và tổ chức.

Những người lập trình viên cần đọc bài này để hiểu cách bảo vệ quyền riêng tư và quyền sở hữu trí tuệ của người dùng trong ứng dụng AI, từ đó xây dựng hệ thống quản lý consent hiệu quả hơn trong các dự án tương tương tương tác với dữ liệu cá nhân.

GDPR at 10: Landmark data protections, increasing business burden

Ten years after GDPR came into force, the regulation has achieved significant data protection milestones — 71% of German companies now comply, up from 7% in 2018, and total fines have exceeded €6 billion. However, business dissatisfaction is growing: 81% of companies say GDPR complicates their processes, and 97% rate compliance effort as high. The biggest emerging tension is with AI development, where 69% of companies say data protection makes it difficult to train AI models with sufficient data. Industry groups like Bitkom are calling for a risk-oriented reform that eases formal burdens while maintaining protections where real risks exist, warning that AI is being developed outside Europe while still being used there — offering neither data protection benefits nor competitive advantages for the EU.

The US government asks OpenAI to slow its next model’s release

The Trump administration has asked OpenAI to stagger the release of GPT-5.6, requiring customer-by-customer access approval during an initial preview period. The request came from the Office of the National Cyber Director and the Office of Science and Technology Policy, framing the concern as cybersecurity-related. This marks the first time the US government has preemptively asked an American AI company to restrict a model launch before it happens. The move follows a similar government directive that pulled Anthropic's most capable models from the market, signaling Washington is now actively shaping frontier AI release schedules rather than reacting after the fact. The arrangement gives a government agency direct influence over who gets early access to a frontier model, shifting control temporarily from the company to the state.

GRC is broken. FedRAMP 20x might fix it

Traditional compliance frameworks like SOC 2 and ISO 27001 have become performative exercises in storytelling rather than genuine security assurance. FedRAMP 20x is pushing a fundamentally different model: machine-readable evidence, APIs, telemetry, and continuous operational visibility instead of point-in-time snapshots and curated screenshots. The author shares their firsthand experience navigating FedRAMP 20x, including entering the moderate pilot without completing the low pathway first, and argues that the discomfort of exposing full operational truth — including bypassed approvals, drift events, and identity lifecycle failures — is precisely the point. The broader GRC engineering movement mirrors this shift, treating compliance as an engineering discipline with telemetry pipelines and continuous assurance rather than annual documentation exercises. The future of trust, the author argues, will be queryable operational data rather than polished PDF certifications.

Why AI and Compliance Are Forcing A New Security Operating Model, with Rapid7's Corey Thomas & Sabeen Malik

Rapid7's Corey Thomas and Sabeen Malik discuss how frontier AI and evolving compliance requirements are forcing a new security operating model. AI is accelerating vulnerability discovery faster than existing standards and governance frameworks can handle, creating noise rather than safety unless paired with exploitability context. Compliance is shifting from annual snapshots to continuous evidence collection. The future model is AI-driven but human-led, connecting security operations, risk management, and compliance through live operational context rather than disconnected reports.

Solo-maxxing: How tech taught Gen Z to monetise being alone

Solo-maxxing — the trend of deliberately maximising time alone and treating singlehood as a lifestyle goal — is examined as a symptom of how technology has commodified solitude. The term traces its roots from 1940s game theory through incel forums to mainstream TikTok culture. Young people are retreating from dating apps due to financial costs and emotional exhaustion, with Tinder losing 7% of paying users in 2024. Algorithms reward aestheticised loneliness content, pushing creators to brand themselves around permanent solitude. The deeper concern is that the same tech ecosystem that gamified intimacy now monetises the retreat from it — and is positioning AI companions as the next step. Regulatory responses like the EU AI Act, Digital Services Act, and US state-level AI companion laws are emerging but are seen as insufficient, since they address transparency rather than the deeper question of what happens to human connection when companionship becomes frictionless and on-demand.