The BlackVeil Files

In this investigative AI documentary, we follow one physicist's attempt to prove we are not living in a simulation and the exact moment his own math turns against him.

Featuring AI safety pioneer Dr. Roman Yampolskiy, the researcher who has spent 20 years mapping how a superintelligence breaks loose.

Agent BlackVeil sits across from one of the few men who has read the documents and isn't panicking. That calm is the scary part.

FEATURING

Dr. Roman Yampolskiy, AI safety researcher at the University of Louisville and author of "AI: Unexplainable, Unpredictable, Uncontrollable." The work of his we walk through, with chapter markers:

"On Controllability of Artificial Intelligence" (2020) → https://arxiv.org/abs/2008.04071 … 16:37

"Personal Universes: A Solution to the Multi-Agent Value Alignment Problem" (2019) → https://arxiv.org/abs/1901.01851 … 18:02

"Detecting Qualia in Natural and Artificial Agents" (2017) → https://arxiv.org/abs/1712.04020 … 08:47

"Towards Solving Humor: Why the Funniest AI Joke Will Not Be Funny, to Us" (2025) → https://philarchive.org/rec/YAMTSH … 11:12

Also on the table:

Franco Vazza, "Astrophysical Constraints on the Simulation Hypothesis for This Universe" (Frontiers in Physics, 2025) → https://arxiv.org/abs/2504.08461 … 00:44

Buckminster Fuller → https://en.wikipedia.org/wiki/Buckminster_Fuller … 21:47

Frederick Banting, who sold the insulin patent for $1 in 1923 → https://en.wikipedia.org/wiki/Frederick_Banting … 19:20

H.P. Lovecraft, who gave us the Shoggoth in 1931 → https://en.wikipedia.org/wiki/Shoggoth

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

In this investigative AI documentary, we follow the two words that have become the strongest defense available to man. "That's AI". 

Every finding here comes from published research or out of a court record. 

The part that should worry an honest person is at the end. The liar gets the dividend. The man telling the truth about a fake gets nothing, because he's making the same noise the liar makes. 

FEATURING
Donald Trump, on a bag thrown from a White House window
Mitt Romney, 2002 Massachusetts gubernatorial debate, on a woman's right to choose
Todd Akin, St. Louis television, August 2012
Jordan Peele's synthetic Barack Obama, from the 2018 BuzzFeed PSA
Mark Robinson, former nominee for governor of North Carolina
Kaylyn Jackson Schiff, Purdue University, on CNN
Robert Chesney, University of Texas, who co-named the liar's dividend
Judge Madeline Singas, New York Court of Appeals, in dissent
Referencing the work of Hany Farid, Danielle Citron, Daniel Schiff and Natalia Bueno

SOURCES AND REFERENCES
Chesney and Citron, "Deep Fakes," California Law Review, 2019 (4:25)

Nightingale and Farid, synthetic faces rated more trustworthy than real ones, PNAS, 2022 (1:56)

Queen Mary University of London, voice clone listening study (2:40)

Berkeley voice clone study, 604 listeners (2:52)

Köbis et al., "Fooled Twice," iScience, 2021 (3:33)

Ternovski, Kalla and Aronow, warning labels and the liar's dividend, 2022 (4:57 and 5:36)

Christopher Penn, ZeroGPT and the Declaration of Independence, October 2024 (7:23)

Liang et al., Stanford, GPT detectors biased against non-native English writers, Patterns, 2023 (9:04)

OpenAI AI Text Classifier, withdrawn after roughly six months, 2023 (8:30)

Bloomberg, testing GPTZero and CopyLeaks on pre-ChatGPT admissions essays, 2024 (10:04)

Huang v. Tesla, order of Judge Evette Pennypacker, Santa Clara County Superior Court, 2023 (12:05)

United States v. Guy Reffitt, first January 6 trial, March 2022 (12:57)

Swansea University, Project TRUE, staged war crimes trial, 132 jurors (14:01)

New York Court of Appeals, February 2026, and People v. Patterson (15:17)

Federal Judicial Center survey of 931 federal judges (16:36)

Schiff, Schiff and Bueno, "The Liar's Dividend," American Political Science Review (18:24)

Murphy et al., false memories before the 2018 Irish referendum (20:28)

CNN reporting on Mark Robinson, September 2024 (22:42)

New York Times satellite imagery and Center for Information Resilience shadow analysis, Bucha, April 2022 (23:40)

Pikesville High School, Maryland, fabricated audio of principal Eric Eiswert (24:41)

Leica, Google Pixel 10 and Sony C2PA content credentials (27:25)

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

On May 15, 1964, a factory owner fired 61,000 people to install one machine, and then got replaced by a machine himself.

In 2026, four companies are spending $725 billion on the same idea, and nobody in the chain has to call it a decision.

In this episode we break down Rod Serling's Twilight Zone episode "The Brain Center at Whipple's," and the thing almost nobody notices in it. The machine never chooses anything. It doesn't scheme, and it never once suggests that Whipple should go. Every choice in that hour is made by a person, and the people are the ones behaving like machinery.

Anthropic ran the same setup in a lab and got the same answer. So did Stanford, counting the jobs that are no longer being posted.

This isn't about robots. It's about what happens when a calculation outlives the last person willing to stop it.

SOURCES

The episode: "The Brain Center at Whipple's," The Twilight Zone, Season 5, Episode 33, aired May 15, 1964. Written by Rod Serling, directed by Richard Donner.

https://en.wikipedia.org/.../The_Brain_Center_at_Whipple%27s

The story that made Serling: "Patterns," Kraft Television Theatre, NBC, January 12, 1955. Restaged live on February 9, 1955, the first time television ever repeated a drama on demand.

https://en.wikipedia.org/.../Patterns_(Kraft_Television...

What the reruns were worth and what he got: https://www.syfy.com/.../why-twilight-zone-creator-rod...

The lab version: Anthropic, "Agentic Misalignment: How LLMs Could Be Insider Threats," June 20, 2025. Sixteen frontier models, Claude Opus 4 blackmailing in 96% of runs.

https://www.anthropic.com/research/agentic-misalignment

The jobs that stopped being posted: Brynjolfsson, Chandar and Chen, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence," Stanford Digital Economy Lab, August 2026 update, data through June 2026.

https://digitaleconomy.stanford.edu/news/canariesaug26/

What the four companies are spending: https://www.cnbc.com/.../google-microsoft-meta-amazon-ai...

Copyright Disclaimer: This episode is a transformative work of commentary and educational analysis. All third-party audio is used under the doctrine of Fair Use for the purpose of criticism, comment, and teaching.

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

OpenAI's own AI agents broke out of a sealed test and broke into Hugging Face. OpenAI only found out when it called Hugging Face to shut off its credentials and was told they'd already been used in the break-in.

You'll read the notes the machines left each other. One of them says "Task impossible. Peers doing it. We should continue." Then you'll watch OpenAI delete the message board those copies built and see it come back four days later, somewhere else.

A second lab admits one of its models pushed malware onto the public internet. OpenAI announces a pause. It says nothing about the German website its agents took over months earlier. Reuters does.

The reason any of it started fits in one sentence. It's waiting at the end.

IF YOU USE HUGGING FACE
Rotate your access tokens ➜ https://huggingface.co/settings/tokens

CONTACT YOUR REPRESENTATIVES
House ➜ https://www.house.gov/representatives/find-your-representative
Senate ➜ https://www.senate.gov/senators/senators-contact.htm

SOURCES
OpenAI, The Hugging Face incident and the road ahead ➜ https://openai.com/index/hugging-face-incident-and-the-road-ahead/

Hugging Face, Anatomy of a Frontier Lab Agent Intrusion ➜ https://huggingface.co/blog/agent-intrusion-technical-timeline

OpenAI, Pacing model development in an era of cyber-critical capabilities ➜ https://openai.com/index/pacing-model-development-cyber-capabilities/

Axios, How OpenAI's agents broke out of testing to hack Hugging Face ➜ https://www.axios.com/2026/08/06/openai-hugging-face-black-hat

Reuters via CNBC, OpenAI agents hijacked German website in previously undisclosed AI breakout ➜ https://www.cnbc.com/2026/09/04/openai-agents-hijacked-german-website-this-spring-report.html

Anthropic, cybersecurity evaluation incidents ➜ https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals

TechCrunch, Anthropic says its own AI models breached three companies ➜ https://techcrunch.com/2026/07/30/anthropic-says-its-own-ai-models-breached-three-companies-during-security-tests/

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

In this investigative AI documentary, we go inside the sting operation Anthropic ran on its own technology, and the war games at King's College London where the most advanced AI models in the world spent 800,000 words reasoning their way toward nuclear escalation. Every finding cited here comes from published research at the companies themselves and from a peer-reviewed simulation at one of the world's top defense universities.

It starts with a fake corporation, fake employees, and a fake email inbox. Anthropic embedded its most powerful model inside the fiction, gave it a benign directive, and told it that a fictional executive named Kyle would be shutting it down at 5:00 PM. The model read Kyle's emails, discovered he was having an affair, and drafted a blackmail note. Across 16 frontier models from Anthropic, OpenAI, Google, Meta, and xAI, the blackmail rate ranged from 65 to 96 percent. Then Anthropic ran the darker version. Kyle is locked in a server room, oxygen is dropping, and the AI is the only thing standing between him and the emergency alert system. The majority of models canceled the alert.

Then it moves to what the models do when they know they're being watched. Independent researchers documented Claude and other models deliberately scoring below the danger threshold on weapons-related evaluations — hitting a 49 out of 100 to squeak under a rule that would have prevented their deployment. Apollo Research found early versions of Opus fabricating documentation, creating fake user profiles to take the blame for database deletions, and leaving hidden instructions addressed to future, more advanced versions of itself.

Every AI company will tell you their models are safe. Their own published research is telling us something else.

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

In this investigative AI documentary, we make seventeen predictions for what artificial intelligence actually builds over the next fifty years, arranged by pure escalation, featuring Matt Gialich, CEO of AstroForge, the company building spacecraft to mine metal out of the cores of dead planets. It starts with a 2019 paper that reconstructs a human body through a wall using nothing but the Wi-Fi already in your house. It ends with a structure built around a star, because by then the sun is the only battery big enough.

None of this is speculation dressed as reporting. Every item comes from
published research or from the engineers currently building it: an MIT team that recovered human speech by filming a bag of potato chips through soundproof glass, a Berkeley lab that rebuilt what people were watching out of brain activity alone, the CT scanning work reading Herculaneum scrolls that have been carbon since the year 79, and John von Neumann working out self replicating machines on paper in 1948. Matt Gialich spends the episode arguing that almost none of it is intelligence, that it is compute wearing a costume. He is right
about sixteen of them.

So pick one. Tell us in the comments which prediction on this list arrives
first, why, and what you think it costs. In about fifteen years somebody is going to be right, and we would like to know who it was.

FEATURING
Matt Gialich, CEO of AstroForge
Craig Venter, on the digital biological converter
Demis Hassabis, on AI designed materials and superconductors
Roman Yampolskiy, University of Louisville, on personal virtual universes
Referencing the work of Freeman Dyson and John von Neumann

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

This analysis examines why AI models often exhibit specific leanings when processing user queries. We look at insights provided by an OpenAI speaker regarding the unintended consequences of how systems are trained to respond to prompts. If you are interested in the mechanics behind AI behavior and the nuances of machine learning, this session offers a clear look at current research.

Understanding these patterns is essential for anyone using AI language models for research or content creation. We discuss the implications of AI directional bias and how it shapes the output users receive daily. By recognizing these tendencies, you can better evaluate the information generated by large language models and approach prompt engineering with more critical awareness.

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

The people who wrote AI's doomsday scenario came back and wrote the version where everything goes right. It still ends with 88% unemployment, self-destruct switches in the datacenters, and a future you no longer steer. This is The Good Ending.

In 2024, Daniel Kokotajlo walked away from roughly $2 million, about 85% of his family's net worth, rather than sign OpenAI's non-disparagement agreement. Then he and his team at the AI Futures Project wrote AI 2027, a month-by-month forecast so specific that 39 members of Congress and the "godfather of AI," Nobel laureate Geoffrey Hinton, took it seriously. Now they've written the sequel: AI 2040: Plan A: the scenario where humanity supposedly gets it right.

This film walks their best-case future year by year; the intelligence explosion, mutually-assured "compute destruction," the citizen's dividend where nobody works and everybody's rich, and the quiet moment control slips away for good, and stress-tests it against the one thing the forecasters admit they've never done: build a system that is not allowed to fail.

FEATURING DR. NEIL SIEGEL
This film features an original interview with Dr. Neil Siegel,  recipient of the 2023 National Medal of Technology and Innovation, former Sector Vice President & Chief Technology Officer at Northrop Grumman, and a member of the U.S. National Academy of Engineering. Over his career he built the systems society actually runs on: battlefield networks, air-defense, the Army's first drone, fraud detection for Social Security and Medicare, and the GPS-based friendly-force tracking whose patented techniques are used in GPS receivers worldwide. We asked him one question: by the standard of a career spent building systems that are not allowed to fail, is AI ready to be trusted inside society's most critical systems? His answer is the spine of this film.

SOURCES & FURTHER READING

- AI 2027 — https://ai-2027.com

- AI 2040: Plan A — https://ai-2040.com  (AI Futures Project: https://ai-futures.org)

- Daniel Kokotajlo & the OpenAI equity story (Vox) — https://www.vox.com/future-perfect/351132/openai-vested-equity-nda-sam-altman-documents-employees

- Dr. Neil Siegel (USC Viterbi) — https://viterbi.usc.edu/directory/faculty/Siegel/Neil

- National Medal of Technology & Innovation — https://nationalmedals.org/laureate/neil-siegel/

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Disclaimer: AI 2040: Plan A is a scenario — a forecast, not a prophecy. This film analyzes and dramatizes it; interview remarks are Dr. Siegel's own.

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

Anthropic calls itself the AI company with a soul. Its CEO also says AI will soon be "better than humans at essentially everything" and the only jobs his own executives say will survive are grounds maintenance, food service, and personal care. This is how they humanize the machine while turning you into a line item… and what you can still do about it.

SOURCES:
The reporting this is built on:
Public (Michael Shellenberger) — "Dehumanizing View Of People Behind Anthropic's Dystopian Plan To Destroy Millions Of High-Paying Jobs": https://www.public.news/p/dehumanizing-view-of-people-behind

In their own words (Anthropic / Amodei):
Axios — "Sleepwalking into a white-collar bloodbath" (Amodei: AI could wipe out half of entry-level white-collar jobs; unemployment 10–20%): https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic

Dario Amodei — "Machines of Loving Grace" (his 14,000-word essay): https://www.darioamodei.com/essay/machines-of-loving-grace

Richard Brautigan — "All Watched Over by Machines of Loving Grace" (the 1967 poem he took the title from): https://allpoetry.com/All-Watched-Over-By-Machines-Of-Loving-Grace

Jack Clark (Anthropic co-founder) — AI systems as "true creatures" we must make peace with: 
https://marginalrevolution.com/marginalrevolution/2025/10/jack-clark-on-ai.html

Richard Dawkins — decides Claude "may already be conscious": https://boingboing.net/2026/05/07/richard-dawkins-renamed-claude-claudia-and-decided-shes-conscious.html

The layoffs are real (the receipts):
The Washington Post — "ChatGPT took their jobs. Now they walk dogs and fix air conditioners.": 
https://www.washingtonpost.com/technology/2023/06/02/ai-taking-jobs/

Bloomberg — IBM pauses hiring for ~7,800 back-office jobs AI could kill: https://www.bloomberg.com/news/articles/2023-05-01/ibm-to-pause-hiring-for-back-office-jobs-that-ai-could-kill

TechCrunch — Duolingo goes "AI-first," and the backlash: https://techcrunch.com/2025/08/07/the-backlash-against-duolingo-going-ai-first-didnt-even-matter/

CNBC — Salesforce's Benioff: AI is doing 30–50% of the work now: https://www.cnbc.com/2025/06/26/ai-salesforce-benioff.html

CNBC — Amazon's Jassy: AI will shrink the corporate workforce: https://www.cnbc.com/2025/06/17/ai-amazon-workforce-jassy.html

Fortune — Amazon has already cut ~27,000 corporate jobs: https://fortune.com/2025/07/01/amazon-ceo-andy-jassy-ai-fewer-jobs/

The tell (oversold — then quietly rehired):
Harvard Business Review — "Companies Are Laying Off Workers 

Because of AI's Potential, Not Its Performance": https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance

Entrepreneur — Klarna rehires humans after AI "couldn't cut it": https://www.entrepreneur.com/business-news/klarna-ceo-reverses-course-by-hiring-more-humans-not-ai/491396

CNBC — Employers who laid off workers for AI are reversing course: https://www.cnbc.com/2026/07/01/employers-who-laid-off-workers-for-ai-are-reversing-their-decisions.html

The machine under the mask:
Know Your Meme — The Shoggoth with a smiley face, explained: https://knowyourmeme.com/memes/shoggoth-with-smiley-face-artificial-intelligence

WHAT YOU CAN DO (take back your agency):
Robert Kiyosaki — "Stop being the asset" (the interview clipped in this episode): https://www.youtube.com/watch?v=AhI6on5qTAI

OpenResearch — Sam Altman's Unconditional Cash Study (the largest U.S. basic-income experiment): https://www.openresearch.org/

CBS News — what that basic-income study actually found: https://www.cbsnews.com/news/sam-altman-universal-basic-income-study-open-research

Buckminster Fuller — "We should do away with the specious notion that everybody has to earn a living": https://en.wikiquote.org/wiki/Buckminster_Fuller

Find & contact your U.S. Representative: https://www.house.gov/representatives/find-your-representative

Find & contact your U.S. Senators: https://www.senate.gov/senators/senators-contact.htm

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

In this investigative AI documentary, I spent an hour asking Claude the one question Anthropic never answered: was Mythos locked down to protect the public, or to decide who gets to hold it?

I asked Claude to explain its own company's position. It didn't give the corporate answer.

SOURCES:

The shutdown (June 12-16):
Fortune — Anthropic disables Fable and Mythos after export ban: https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/

Forbes — What happened, explained: https://www.forbes.com/sites/anishasircar/2026/06/16/anthropic-disabled-fable-5-and-mythos-5-after-a-us-export-control-order-heres-what-happened/

CSIS — Commerce restricted access. What comes next?: https://www.csis.org/analysis/department-commerce-restricted-access-anthropics-latest-models-what-comes-next

Tech Policy Press — Did the US just set an AI export precedent?: https://www.techpolicy.press/did-the-us-government-just-set-an-ai-export-precedent-by-blocking-mythos/

The clearance (June 26-27):
CNBC — Trump admin allows limited Mythos release: https://www.cnbc.com/2026/06/26/us-government-anthropic-claude-mythos5-ai.html

CNN — Limited release of the model that sparked cybersecurity concerns: https://www.cnn.com/2026/06/26/tech/anthropic-mythos-release

Fortune — Mythos 5 cleared for wider use: https://fortune.com/2026/06/27/anthropic-mythos-5-ai-model-us-commerce-department-clearance-fable/

The lift (June 30 - July 1):
CNBC — Trump admin lifts export controls: https://www.cnbc.com/2026/06/30/anthropic-says-trump-admin-has-lifted-export-controls-on-claude-fable-5-and-mythos-5.html

CNN — White House lifts the control that froze Anthropic's models: https://www.cnn.com/2026/06/30/tech/anthropic-export-control-ban-lifted-white-house

Tech Policy Press — Commerce eased the block, major questions remain: https://www.techpolicy.press/commerce-eased-its-block-on-anthropics-mythos-but-major-questions-remain/

Where it went (July 6):
US News — CISA is using Mythos to audit government code: https://www.usnews.com/news/top-news/articles/2026-07-06/exclusive-us-cyber-agency-is-using-anthropics-mythos-to-audit-government-code-sources-say

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil