Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: robotics/AI – Page 30

The silent data leak hidden inside encrypted reasoning traces of frontier AI models

To hide this reasoning while maintaining conversational context across multiple API turns, major providers such as Anthropic and OpenAI use Authenticated Encryption with Associated Data (AEAD) envelopes. AEAD is a cryptographic standard that encrypts data while attaching metadata to verify that the payload has not been modified in transit.

Providers designed these encrypted envelopes around three goals:

1. Confidentiality: Prevent clients from reading the intermediate reasoning tokens to protect proprietary information.

The dangers of trusting blackbox machine learning

The success of deep learning in the past decade has increased interest in the field of artificial intelligence. But the rising popularity of AI has also highlighted some of the key problems of the field, including the “black box problem,” the challenge of making sense of the way complex machine learning algorithms make decisions. The Apple Card disaster is one of many manifestations of the black-box problem coming to light in the past years.

The increased attention to black-box machine learning has given rise to a body of research on explainable AI. And a lot of the work done in the field involves developing techniques that try to explain the decision made by a machine learning algorithm without breaking open the black box. But explaining AI decisions after they happen can have dangerous implications, argues Cynthia Rudin, professor of computer science at Duke University, in a paper published in the Nature Machine Intelligence journal.

“Rather than trying to create models that are inherently interpretable, there has been a recent explosion of work on ‘explainable ML’, where a second (post hoc) model is created to explain the first black box model. This is problematic. Explanations are often not reliable,” Rudin writes. and can be misleading, as we discuss below.

Moving beyond passive RAG: How to implement active memory reconstruction for AI agents

To see how this works in practice, consider a concrete enterprise use case: a customer success agent that needs to answer, “Why was this customer promised a different renewal price, and should we honor it?”

The answer may require reconstructing a causal chain across a support conversation, a sales exception, a contract clause, a billing-system update, and a later internal note. Traditional similarity-based RAG might retrieve the most recent billing document or the most semantically similar support ticket, but fail to connect the causal chain.

With MRAgent’s active reconstruction, the agent can extract cues like the customer’s name, follow associative tags to the original support ticket, retrieve the sales exception, and link it to the billing update. It navigates these semantic relationships, prunes irrelevant branches, and iteratively stops once it has enough evidence to answer the query.

US authorities say Siemens controllers used for water and other infrastructure are being targeted by hackers threat actors use AI tools to generate exploitation scripts

Attacks on these industrial controllers could lead to sabotage of critical infrastructure.

Longevity Day

LONGEVITY DAY — a science-fiction trailer for the Future Vision XPRIZE.

In 2035, an eighty-eight-year-old biologist with cancer and failing memory fights to restart the AI-assisted rejuvenation research that may be her last chance — three years after a catastrophic AI failure nearly convinced the world to abandon frontier research altogether.

A story about making aging optional.

Written, directed and produced by Sergio M.L. Tarrero.
Alianza Futurista • https://alianzafuturista.org.

Submitted to the Future Vision XPRIZE, presented by the XPRIZE Foundation, Google and Range Media Partners.
https://futurevisionxprize.com.

#FutureVisionXPRIZE

Continue reading “Longevity Day” | >

Cloudflare Just Wired 20% Of The Web For AI Agent Payments

Cloudflare, which routes roughly 20% of global web traffic, launched Cloudflare Wallets and the x402 protocol on August 4, 2026, giving AI agents a funded, capped way to pay for APIs and data without a human clicking checkout. More than 20 companies are already participating in these agent-initiated payment flows. The infrastructure isn’t fully live yet, but the spend-governance policy question is already overdue.

AI agent payments just moved from a theoretical problem to live infrastructure at internet scale. During a release cycle it called Agents Week, Cloudflare shipped Cloudflare Wallets alongside the x402 protocol, reviving HTTP status code 402, Payment Required, a code that has sat unused in the web standard for roughly thirty years, according to an August 2026 technical breakdown. A server that wants payment now returns a 402 response with machine-readable terms; an AI agent attaches proof of payment in stablecoins and the transaction settles without a checkout page, a login, or a human in the loop.

Cloudflare processes approximately 20% of all web traffic, so when the company builds a payment primitive directly into that infrastructure, it isn’t a startup experiment, it’s plumbing, according to Forkast’s coverage of the launch. The design splits custody from spending: an Account Wallet belongs to a person or company and holds real funds, while Virtual Wallets are capped, delegated allowances handed to individual AI agents, functioning like a corporate checking account issuing restricted debit cards to employees, except the employees are software. AI agent payments under this model inherit spend limits set by a human owner, not by the agent itself.

New Technique Could Slash AI’s Memory Energy Use by Thousands of Times

The microscopic magnetic flips behind digital memory could soon use thousands of times less energy, offering a new way to shrink AI’s rapidly growing power footprint.

Artificial intelligence is creating and processing data on an enormous scale. Searches, recommendations, generated images, scientific simulations, and large language models all depend on information that must be repeatedly stored, transferred, retrieved, and rewritten. Each operation consumes energy, and those costs multiply across the billions of devices and sprawling data centers that support the digital world.

Researchers at the University of Edinburgh have now developed a mathematical framework designed to slash the energy required to write information in future magnetic memory. Rather than creating a new memory material, the approach changes how the magnetic state representing a digital bit is flipped.

/* */