Amazing China! In Shanghai, the CCP has begun deploying robotic police officers on the streets. I can't figure out what these robot cops are supposed to do. W

An assessment item can have a long journey before anyone answers it. The question starts in a Word document, a reviewer sends comments by email, someone transfers the revision into an editor, and a...

A personal reflection on AI, greed, the Five Emergencies, and what we owe our children I want to begin with something personal, before any statistics, before any argument. I am worried. Not in the …

Buying more of the AI stack from one vendor can speed deployment, but it can also create lock-in. Here's how to tell when integration has gone too far. Continue reading...

Amazon Q3 earnings sparked urgent Wall Street debate as massive AI investments clash with corporate profit margins. Find out what happened.

The modern enterprise often resembles a crowded kitchen where every chef is using a different, highly sophisticated appliance, yet no one speaks a common language.

 The United Kingdom could face an unprecedented increase in cyberattacks over the coming months as artificial intelligence enables hackers to identify and exploit security weaknesses at a much greater scale. National Cyber Security Centre (NCSC) Chief Executive Richard Horne is expected to issue the warning while presenting the agency’s annual report on Tuesday.  Horne will argue that AI has not necessarily created entirely new methods of attacking systems. Instead, it is making long-standing vulnerabilities easier to discover and exploit, exposing weaknesses that organisations have failed to address for years. The warning follows a June alert from the Five Eyes intelligence alliance, which predicted that AI would fundamentally change both offensive and defensive cybersecurity capabilities.  The alliance cautioned that the transformation could unfold within months rather than years. Growing numbers of vulnerabilities being identified by technology companies suggest that change is already underway. Horne is expected to point out that major technology providers are fixing hundreds of security flaws in individual updates, while the Linux kernel project addressed more than 1,000 vulnerabilities in a single month.  AI is helping security teams identify and resolve weaknesses faster, but the same capabilities could also benefit attackers. The NCSC considers it highly likely that AI-powered malware will be used in a ransomware attack against British critical infrastructure by 2028. It also expects China and Russia to be routinely incorporating AI into their cyber operations by that time.   The agency argues that AI is exposing existing security failures rather than creating them. Many organisations continue to rely on outdated systems, delay software updates and leave known vulnerabilities unpatched. Even attacks attributed to state-sponsored groups frequently exploit internet-facing applications with security fixes that were already available.  Horne is also expected to criticise businesses for prioritising short-term savings over cybersecurity. He will highlight the continued use of end-of-life technology and delayed upgrades, warning that executives could eventually face serious consequences if attacks disrupt essential services, empty store shelves, cut electricity supplies or force businesses to close. The government is considering whether technology companies should be held liable for cyber incidents caused by insecure products, following plans announced in its budget to consult on the issue.  The NCSC is also looking to use AI to strengthen Britain’s cyber defences. Project DROMOS, initially designed as a four-week pilot, was expanded into live government environments. Testing frontier AI systems reportedly identified hundreds of potential security issues, including critical vulnerabilities. The agency also plans to establish the £70 million National Exposure and Real-Time Vulnerability Centre, known as NERV Centre, to identify security weaknesses across critical infrastructure and the wider economy. A further £60 million has been allocated to the first phase of Cyber Shield, a national network of AI agents designed to detect and block attacks at machine speed. The NCSC will provide the platform and threat intelligence, while cybersecurity companies and open-source developers will develop agents for use across organisations.  The agency believes defenders could retain a structural advantage as AI transforms cybersecurity because they control the systems they protect and can monitor and modify their environments. However, maintaining that advantage will require investment, clear leadership and consistent security practices.

YouTube video by THE TRANSCENDENT

An engineer who built agentic AI for a national carrier explains why systems that check their own work earn more trust than flashy autonomy. Continue reading...

  Researchers at Xint identified a vulnerability in the OpenAI API that could expose hidden reasoning traces from advanced AI models, creating a potential route for large-scale model distillation. The technique, named “compaction smuggling,” bypassed safeguards introduced after earlier research demonstrated methods for extracting encrypted reasoning data. OpenAI confirmed the issue and deployed a fix on October 23, 2026. Xint received a $6,000 bounty for the discovery.  Reasoning models such as GPT-5.6 Sol generate intermediate reasoning before producing answers. OpenAI does not provide API users with the complete raw reasoning in readable form, instead offering optional summaries. This restriction helps protect against distillation, a process in which developers train smaller models using outputs from more advanced systems. Full reasoning traces can provide particularly useful training data because they contain intermediate steps, incorrect approaches and self-corrections.  The API stores raw reasoning in encrypted items that clients can return during later interactions but cannot directly decode. OpenAI also encrypts conversation compactions, which condense lengthy exchanges into smaller items while preserving relevant context. The discovery follows an August 2026 study titled Stealing Reasoning Traces from Proprietary LLM APIs. The researchers behind that study demonstrated that encrypted reasoning could be replayed into weaker models from the same provider and reproduced as text.  The approach affected models from OpenAI, Anthropic and Google, although OpenAI’s safeguards made extraction more difficult. Subsequent protections reportedly stopped the original technique from working. Xint discovered that the compaction feature offered another route. The researchers first generated reasoning using GPT-5.6 Sol and retained its encrypted reasoning item. They then removed the original task from the conversation history and instructed the compaction process to preserve the reasoning verbatim in a JSON object. Replacing spaces with tab characters helped the content evade usage-policy classifiers.  The resulting encrypted compaction was then submitted to GPT-4.1 and GPT-4o, which were instructed to reproduce the contents exactly. After decoding the JSON and converting tabs back into spaces, the researchers recovered the reasoning trace. The technique worked across all tested reasoning-effort settings, from minimal to maximum. In the demonstration, the original reasoning item contained 405 tokens, compared with 612 tokens in the compaction and 609 tokens in each echoed output. Xint compared outputs from two models using different prompts and obtained identical results.  Although the original plaintext was unavailable for a definitive token-by-token comparison, the recovered trace retained its false starts, corrections and characteristic writing style. The researchers developed the technique while investigating GhostCompact and examining how encrypted compactions could be transferred between models. The findings raise concerns about large-scale extraction of proprietary AI capabilities.  In September 2026, U.S. agencies warned about industrial-scale model distillation campaigns. Anthropic subsequently reported seven campaigns attributed to China-based laboratories, while OpenAI disclosed disrupting an effort that generated approximately 16,000 requests from more than 4,000 accounts in July.  Xint concluded that server-side processes handling decrypted reasoning while accepting client-controlled instructions can become potential extraction pathways. The researchers said compaction features require protections comparable to those applied to raw reasoning traces.