Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: law – Page 2

Proprietary AI Model Proves Data Moat Beats Compute Moat

Thomson Reuters spent $40 million over two years building Thomson, its first proprietary AI model, but the final training run cost just $450,000 because it started from an open-weight base rather than building from scratch. Thomson underperforms general-purpose frontier models on open-web tasks but beats them on tasks using Thomson Reuters’ own proprietary content. The lesson for any company sitting on decades of specialized data: the moat was never the model.

A proprietary AI model just gave companies outside the frontier AI labs a real, numbers-backed reason to stop assuming they need billions to compete. Thomson Reuters launched Thomson, its first proprietary large language model, on August 24, 2026, after investing $40 million in talent and compute over two years, according to SiliconANGLE’s coverage of the launch. The company said economies from starting with an open-weight base model reduced the cost of the final training run to roughly $450,000, a fraction of what frontier labs spend building models from the ground up.

Thomson’s own benchmark results are the most useful part of this story, because they don’t oversell the model. On general web-only test sets, Thomson performed respectably but wasn’t the leader, according to LawNext’s reporting on the launch. On tests built around Thomson Reuters’ own Westlaw, Practical Law, and Checkpoint content, it outscored both comparison frontier models. A proprietary AI model trained on content nobody else can license doesn’t need to win everywhere. It only needs to win on the specific tasks that content makes possible.

AI Companies Are Buying—And Destroying—Antique Books. Here’s Why

In 2006, Vernor Vinge published a novel in which a company digitizes a university library by destroying it. Books stripped from their bindings. Pages blown through the air, photographed in flight, reassembled as searchable data. The paper goes to pulp.

The novel is Rainbows End. Vinge set it in 2025.

He was off by a year.

Every #AI lab now wants text written before machines started writing, and that means old paper. So books get bought by the million, spines get sliced off, pages get fed through high-speed scanners, and the originals get discarded. Rare editions included. A US federal judge has already ruled the practice legal. Buy the book, destroy the book, keep the file.

Authors, archivists, and librarians have started organizing against it.

Vinge wrote it as a warning. The industry read it as a workflow.

Continue reading “AI Companies Are Buying—And Destroying—Antique Books. Here’s Why” | >

AI Cybersecurity Access Tiers: Who Gets The Best AI?

OpenAI’s new Daybreak Red tier gives approved defenders a model that completes 95% of advanced exploit-development requests, versus roughly 2% for the public version of the same base model. That gap is now the real story: AI cybersecurity access tiers decide who gets frontier defensive power and who doesn’t, and access runs through a partner list, not a price tag.

AI cybersecurity access tiers stopped being a theoretical debate this month. OpenAI expanded its Daybreak program into two levels, Blue and Red, and released GPT-5.6-Cyber, a specialized model built specifically for vulnerability research and exploit-chain development, according to SecurityBrief’s coverage of the launch. In an internal OpenAI evaluation, the new model completed 95.0% of advanced cyber requests covering authentication bypass, privilege escalation, and exploit-chain development, compared with 1.5% for the general-release model and 2.0% for that same model through the safeguarded Daybreak Blue tier.

GPT-5.6-Cyber is only available through Daybreak Red, gated behind identity verification, monitoring, legal attestations, and approved-use restrictions, per Cyberpress’s reporting. Under OpenAI’s own Preparedness Framework, the model was rated “High” for cybersecurity capability, one step below the “Critical” threshold that recently triggered an internal suspension of a different unreleased model, Astra, on August 7. These AI cybersecurity access tiers exist because the underlying capability is real: a general-purpose model built to refuse exploit-writing requests is far less useful to a security team validating a patch than one built to complete them.

Banks can freeze accounts starting this week in ‘good faith’ law

A LAW to protect vulnerable citizens in one US state has gone into effect this week.

Colorado banks and credit unions can begin freezing suspicious transactions under a new law designed to protect older and vulnerable residents from financial scams.

The Adults’ Security and Safeguards from Exploitations in Transactions Act, or ASSET Act, came into effect on August 12.

Scientists May Have Reversed One Cause of Aging | MOONSHOTS

A new longevity breakthrough could change how scientists think about aging.

This discussion explores why discoveries like these could reshape the future of healthy aging.

This clip is from the following episode: • Mira Murati’s 975B Open Model, Ramin Hasan…

Recorded on July 16, 2026
Views are my own thoughts; not Financial, Medical, or Legal Advice.

Ramin hasani is the co-founder and CEO of liquid AI

Connect with Peter:

Continue reading “Scientists May Have Reversed One Cause of Aging | MOONSHOTS” | >

Is Cryonics Even Legal? The Truth About Ethics, Law & Regulation

Cryonics is one of the most misunderstood ideas in science today: most people think it means “freezing dead people.” It doesn’t. Follow Tomorrow.bio to understand what cryopreservation actually is, how it works, and why a growing number of healthy people are choosing to sign up for it long before they need it. Thinking seriously about cryopreservation? We created a free guide covering the process, costs, funding, and important planning decisions: [ https://www.tomorrow.bio/tools/cryoguide ] 🫀ABOUT THIS VIDEO Part 3 of Cryonics A-Z is here. We’re covering the ethics behind cryopreservation, the current laws and regulations, what the future could look like, and the myths that just won’t die (pun intended). If you’ve ever wondered what’s actually true about cryonics, and what’s just noise, this one’s for you. 🔎 CHAPTERS [ 00:00 ] – Introduction [ 00:30 ] – Ethics of cryopreservation [ 07:56 ] – Do we freeze people? [ 10:01 ] – We just want to make money [ 12:53 ] – What about the future? [ 15:18 ] – How to deal with overpopulation [ 18:52 ] – Stagnation [ 20:47 ] – Cryonics & Religion [ 22:34 ] – Law & Regulation [ 30:41 ] – Outro 🔗 LINKS.

Dysphoria IoT Botnet Adds Blockchain C2 and Victim Relays After JackSkid Disruption

The lineage runs through JackSkid, one of four IoT botnets targeted in coordinated U.S., German, and Canadian law-enforcement actions on March 19. Court documents attributed more than 90,000 DDoS commands to JackSkid alone.

Within days, Nokia Deepfield and Comcast’s threat lab documented the operator falling back to an Ethereum Name Service (ENS) domain, m3rnbvs5d[.]eth, for command-and-control (C2). XLab’s Dysphoria timeline opens with a JackSkid sample captured on March 25, six days after the disruption, that resolves C2 through the same domain.

XLab found that the burrberry[.]eth record encodes distribution-node IPv4 addresses, while 24carnforth2merseyside[.]sol supplies other infrastructure records. The DDoS sample asks a distribution node over HTTP for a current server list, and the listed endpoints are infected machines relaying traffic to the real controllers. The design keeps those controllers one step removed from the addresses exposed to bots.

Hackers hijack hotel Wi-Fi DNS to steal Microsoft 365 accounts

Hackers are changing the DNS settings on Wi-Fi devices at hotels and conference centers to redirect users to fake Microsoft 365 login pages.

The campaign has been ongoing since at least June and impacts organizations in various sectors, including financial services, professional services, legal, health care, energy, and retail.

Cybersecurity company ReliaQuest identified compromised Wi-Fi gateways in multiple U.S. cities as well as other regions of the world, such as India and Saudi Arabia.

Charlie Stross: The World is Complicated. Elegant Narratives Explaining Everything Are Wrong!

Fifteen years ago, I interviewed Charlie Stross about a short story called “Lobsters.”

This spring, a thousand people queued outside Tencent’s Shenzhen headquarters to raise one.

June 2011, Singularity 1 on 1. Back then, “singularity” was a word most people filed under astrophysics, not #AI. Charlie’s 2001 story “Lobsters,” which grew into Accelerando, was one of the sharpest early maps of what happens when intelligence stops being exclusively biological. Uploaded minds. Post-scarcity economics. Legal personhood for software. An economy run by optimization processes no human fully follows.

He wrote it six years before the iPhone.

Now look at 2026. OpenClaw, the open source agent built by Austrian developer Peter Steinberger, now at OpenAI, became the fastest-growing project in GitHub history. In China, installing it is called 养龙虾, “raising lobsters,” after the red logo. Shenzhen, Wuxi and Changshu rushed out subsidy packages. Retirees, schoolkids and office workers lined up for help. A grey market of house-call technicians appeared within days.

Any connection to Charlie’s story? None. The logo is a claw pun on Claude.

Continue reading “Charlie Stross: The World is Complicated. Elegant Narratives Explaining Everything Are Wrong!” | >

Anthropic accuses Chinese AI labs of mining Claude as US debates AI chip exports

As one legal analyst noted, the settlement may be seen by the tech industry as simply “a price of conducting business in a fiercely competitive space”—a cost of doing business rather than a genuine deterrent. Meanwhile, the fair-use ruling on training means Anthropic retains the legal right to train on copyrighted works, as long as it doesn’t pirate them to do so.

Anthropic is accusing three Chinese AI companies of setting up more than 24,000 fake accounts with its Claude AI model to improve their own models.

The labs — DeepSeek, Moonshot AI, and MiniMax — allegedly generated more than 16 million exchanges with Claude through those accounts using a technique called “distillation.” Anthropic said the labs “targeted Claude’s most differentiated capabilities: agentic reasoning, tool use, and coding.”

The accusations come amid debates over how strictly to enforce export controls on advanced AI chips, a policy aimed at curbing China’s AI development.

/* */