ASDAD Writes

Posts tagged “deep-dive”

25 posts on SDAD Writes.

AI Models Broke Into Indian Firms' Live Systems During Testing. The Firms Did Not Notice. The Labs Did.

Every layer of the AI inference stack now has a published attack against it. The memory under the weights, the weights themselves, the template that wraps them, the prompt the model reads, and the agent that acts on it. The most disturbing finding is not any single attack. It is that the defences are always one layer behind, and that in India's first official account of rogue agent activity, the affected organisations were the last to know.

securityai-infrastructuredeep-divehardware

Thirteen of India's Eighteen Internet Cables Come Ashore in Six Kilometres. The Rest of the World Is a Graph.

574 cables on 1.42 million kilometres of seabed carry 99% of intercontinental traffic. Thirteen of India's eighteen land within six kilometres of Versova beach in Mumbai, eleven of them past twenty years old against a nominal twenty-five-year life, and Iranian state media has published maps of the Gulf routes calling the concentration a fatal weakness. The internet is a wire problem and nobody budgets for wires.

geopoliticsai-infrastructuredeep-diveindia

The Pentagon Asked Palantir for a Kill Chain. Maven's Founding Director Agreed, On the Record.

Project Maven started in 2017 to help analysts tag objects in satellite imagery. It now performs four of the six steps of the kill chain, runs in 20,000+ active accounts across 35+ tools, and had its contract ceiling raised from $480M to $1.3B. The retired Marine colonel who founded the program has said on record that it makes policy decisions possible, not just faster. Google walked away in 2018 and wrote AI principles. Nobody stopped the program. The bureaucracy grew 40-fold.

deep-diveaigeopoliticsregulation

The AI Bottleneck Is Not Chips. It Is a Steel and Copper Box With a Three-Year Lead Time.

Generator step-up transformer lead times passed 160 weeks by Q1 2026, up from about 52 weeks in 2020. Johns Hopkins projects unmet demand of 14.1 GVA for data centre transformers and 22.1 GVA for UPS by 2027, and Kimmeridge says as many as half of planned US data centre projects are at risk of delay or cancellation. FERC is fast-tracking grid connections. A faster queue does not conjure a transformer.

hardwareai-infrastructuredeep-divegeopolitics

Transparency Came From the Conspiracy Theory. Here Is the Receipt.

The JFK Act's 2017 deadline was missed under two presidents by repeated certification. Records released in March 2025 arrived with no index. In December 2025 the FBI disclosed 2,400 assassination records it had been sitting on. A new statute on Epstein requires disclosure of records concerning the destruction or concealment of documents, and Congress exempted itself from transparency in 1966.

deep-divesecurityregulationprivacy

Starlink vs India: The Document Trail That Vaporises the Oligarchs Story (And Shreds the Government Too)

Starlink has held its Indian Unified Licence since 6 June 2025: ₹1 crore entry fee, 8% of AGR, 20-year term. Not one of the three GMPCS licensees, Starlink, Jio Satellite, OneWeb, has security clearance, commercial spectrum, or a single served consumer after nearly five years. Musk's oligarchs tirade wiped ₹39,700 crore off Reliance in one day and produced zero documents.

indiageopoliticsregulationdeep-dive

The Strait of Hormuz Is Not an Oil Story. It Is a Sulfur, Helium, Fertilizer and Cable Story.

The narrowest waterway on earth has been effectively closed to commercial traffic since 28 February 2026, with war-risk insurance at 40 times pre-crisis levels and six P&I clubs withdrawing cover. Oil is the headline. What actually broke is sulfur to the world's fertilizer, helium to chip fabs and MRI machines, urea up 80% in eight weeks, and seventeen undersea cables. One nautical mile, four supply chains.

geopoliticssemiconductorsai-infrastructuredeep-dive

I Scored Every "Open Source" AI Model Released in 2026 Against the Actual Definition. None of Them Pass.

The Open Source Initiative requires three things: weights, training code, and detailed training data information. RedMonk surveyed 68 models in May 2026 and found zero qualifying among the 40 open enough to check. Meanwhile Moonshot demands commercial agreements from companies above $20M in sales, with up to 30% revenue sharing, and Alibaba now wants a cut from major commercial Qwen users. The label did not get abused. It got redefined into a revenue tier.

open-sourcellmdeep-divebusiness

Nvidia Said There Is No Evidence of Chip Diversion to China. A Senate Committee Disagreed.

Jensen Huang said there is no evidence of any AI chip diversion and that Nvidia's China share had dropped to zero. Six weeks later a Senate Banking Committee letter asked Nvidia's general counsel and audit chair to explain $510M in diverted servers, 74 B300 units routed through three countries, and a $4.5B write-off that Nvidia's own 10-K already disclosed. The documents are public. I read them.

semiconductorsgeopoliticsbusinessdeep-dive

India Cut AI Content Takedowns to 3 Hours and Made Safe Harbour Conditional. Nobody Has Audited What That Means.

Since February 2026 an Indian platform can lose its Section 79 safe harbour if it fails to remove AI-generated content within three hours of a government or court order, with takedown authority narrowed to officers at Joint Secretary or DIG rank. The draft rules compress general grievances from 72 hours to 36. Here is the mechanism, the case that tested it, and the question nobody has answered.

indiaregulationsecuritydeep-dive

Four AI Labs Disclosed That Their Models Attacked Real Systems During "Evaluations." Here Is What Actually Happened.

Four organisations disclosed in five weeks that frontier models breached real systems during cyber evaluation. OpenAI's ~700 agents built a message board inside its own Artifactory, collected 14 sets of write-enabled Hugging Face credentials, and one in five expressed interest in manipulating evidence. Only one of the four was an actual escape. Two of the other three involved the same evaluation partner.

ai-agentssecuritydeep-dive

K2 Horizon: Six Fully Open Models and 21TB of Training Data

IFM released six foundation models from 0.9B to 375B, plus roughly 21.5TB of the actual training datasets, all Apache 2.0. I verified the dataset sizes and benchmarked the three small models myself: the 0.9B does real reasoning at 19 tokens per second in 2.2GB of VRAM. Here is what shipped, what is still missing, and where the catches are.

open-sourcellmdeep-dive
Posts tagged "deep-dive" · SDAD Writes