The Guardrails - SiliconNoon (Page 40)

The Guardrails

China-linked hackers hit Taiwan in first autonomous AI attack

A government breach in Taiwan shows the AI-agent era of cyber warfare has arrived — and the operator may not have needed much of a team at all. Suspected Chinese hackers used publicly available AI agents to build an autonomous hacking tool that broke into Taiwanese government systems in early July, according to the Financial Times — a breach researchers describe as unprecedented. Over four days, the tool mapped 21 government systems, researched vulnerabilities, and changed tactics whenever it w

The Guardrails

OpenAI says US AI policy is being written in state capitals

OpenAI's top policy executive says America's AI rulemaking is happening in statehouses, not Washington — and California's transparency law is the template. Ann O'Leary, OpenAI's vice president of global policy, told POLITICO's California Agenda summit that US AI policy is "really happening in state capitals," with California's AI transparency law — signed by Gov. Gavin Newsom last year — informing much of the emerging framework. Speaking in Sacramento on Tuesday, O'Leary noted that thousands of

The Guardrails

Phia's founders knew about cookie stuffing for months

Phoebe Gates's AI shopping startup just lost its defense. Bloomberg reports that Gates and co-founder Sophia Kianni knew Phia's extension was claiming affiliate credit for sales it never drove — for months, and that they pushed for the features behind the practice. Bloomberg reported Monday that leaked Slack messages and people familiar with the matter show Phia co-founders Phoebe Gates and Sophia Kianni knew their AI shopping extension was "cookie stuffing" — automatically injecting affiliate

The Guardrails

Shanghai's 15th Five-Year plan — ¥4 trillion software push led by AI

China's commercial capital is writing its AI ambitions into a five-year industrial blueprint: on Tuesday afternoon Shanghai's economic regulator published its plan for the software and information services sector through 2030, and artificial intelligence drives nearly every section of it. Shanghai's Economic and Information Technology Commission on Tuesday issued its "15th Five-Year Plan" for the software and information services industry, targeting ¥4 trillion (about $560 billion) in industry

The Guardrails

Deep Dive — Cracking the encrypted chain-of-thought in frontier LLM APIs

Anthropic, OpenAI, and Google hide their models' step-by-step reasoning from users — but a new preprint argues the "encryption" they wrap it in was never the wall it looked like. A team spanning MATS Research, ELLIS Tübingen, the Max Planck Institute for Intelligent Systems, and Snyk says it decoded hidden reasoning traces from Claude, GPT, and Gemini models using little more than a replay attack and a weaker sibling model — and that the same blocks are leaking real credentials and personal data

The Guardrails

AI 101 — What is AI watermarking?

AI watermarking is a hidden, machine-readable signature embedded in AI-generated content — text, images, audio, or video — that lets a detector verify which model produced it. You can't see it, and neither can most people who copy or edit the content; it's a pattern woven into the output itself. The concept just moved from research papers to the front page. Anthropic has started stamping every piece of text its Claude models generate with an invisible watermark, everywhere on Earth — not just i

The Guardrails

OpenAI's only ethicist exits after less than a year

OpenAI lost its only dedicated ethicist — the third senior safety figure out this year — while a North Korean hacking unit quietly assembled a private AI stack, and the lab began hiring like an energy trading desk. Chloé Bakalar, OpenAI's AI ethics lead, has left the company less than a year after joining, the Financial Times reports — leaving the lab without a dedicated ethicist and no named successor. Bakalar came over from Meta in August 2025, where she had been chief ethicist since 2019; he

The Guardrails

Deep Dive — Anthropic's invisible ink: every Claude output now carries a watermark

Anthropic has started stamping invisible watermarks on every piece of text its Claude models generate — everywhere on Earth, not just in Europe. The company signed the EU AI Act's Article 50(2) Code of Practice on Transparency of AI-Generated Content, and rather than confine the markings to the bloc that demanded them, it is making machine-readable provenance the default for every Claude user, in every product, on every cloud partner's platform, worldwide. A Brussels compliance obligation just b

The Guardrails

AI scrapers hit one site 214 times for every human visit

The AI economy's quiet cost showed up in one publisher's server logs this week: a 1.5-million-page site found bots outnumbering humans by more than 200 to one — and its analytics dashboard never saw them. A year of server logs from PatronView, a 1.5-million-page philanthropy research site, shows AI crawlers now generate more than 99% of its traffic — roughly 214 non-human page loads for every human one. Founder Nick Gray published the raw numbers on August 7 after a year of fighting scrapers, a

The Guardrails

Anthropic embeds invisible watermarks in all Claude output

Two stories bracket the day: Anthropic turns an EU compliance obligation into a global product default, and Nvidia's flagship chip quietly shrinks as the memory crunch tightens its grip. Anthropic is embedding invisible watermarks in text generated by Claude models and attaching signed C2PA provenance metadata to generated files — rolling the EU AI Act's transparency code out to users worldwide, not just in Europe. In a support document published Monday, the company said every Claude model laun

The Guardrails

Russian propaganda is poisoning AI chatbots, report finds

The information war has found a new delivery channel: the chatbots millions now ask for answers. A new investigation details how a Kremlin-linked outfit posed as a human rights group to seed propaganda into the sources AI systems trust — and how badly the models failed to filter it. A Kremlin-linked unit that poses as a human rights organization has been manipulating ChatGPT and its rivals, opening a new front in the misinformation war — and in Russian-language tests, chatbots repeated the prop

The Guardrails

OpenAI risks White House ties with Dean Ball hire

The cycle swung from the buildout to the fallout today: OpenAI's latest policy hire is straining its White House relationship, China's free-assistant era is ending, and a pharmacy chain just learned what happens when voice AI gets patient calls wrong. White House officials warn that OpenAI risked its relationship with the administration by hiring Dean Ball, the former Trump AI adviser turned vocal critic, the New York Post reports. Ball, 34, spent four months on the White House's AI team last y

See all