YouTube
Black Hat USA 2026 | ChatMate: Remote Prompt Execution on AI Assistants through Sandbox Escaping
Imagine a user asks an LLM a question about a document. An attacker then gains an interactive prompt on the user's chat session, enabling the attacker to instruct the AI assistant to take actions on behalf of the victim.
In this Briefing, we will reveal…
In this Briefing, we will reveal…
🚨 One prompt injection became a full escape from Microsoft Copilot's sandbox, and it paid $48,000
Researcher Ori's Black Hat talk walks through CVE-2026-32193, now fixed by MSRC. The chain:
• Indirect prompt injection, then code exec and privesc inside the sandbox
• Arbitrary file write in an Azure internal service, TOML injection, /etc/ld.so.preload
• Escape to interactive access to Copilot itself
• The daemon is unauthenticated on localhost on AKS and Container Apps nodes with image streaming, so any SSRF there means root on the node
Researcher Ori's Black Hat talk walks through CVE-2026-32193, now fixed by MSRC. The chain:
• Indirect prompt injection, then code exec and privesc inside the sandbox
• Arbitrary file write in an Azure internal service, TOML injection, /etc/ld.so.preload
• Escape to interactive access to Copilot itself
• The daemon is unauthenticated on localhost on AKS and Container Apps nodes with image streaming, so any SSRF there means root on the node
❤2
🧠 Raschka's new explainer covers how LLMs know word order at all
Sebastian Raschka published a walkthrough of positional embeddings, from GPT-2's learned absolute vectors to RoPE, which MiniMax-M3 applies inside every attention block, with visuals, a little math and a small PyTorch implementation. Section 6 maps which choices 2026 open-weight models use, including some that skip explicit positional encoding entirely.
Sebastian Raschka published a walkthrough of positional embeddings, from GPT-2's learned absolute vectors to RoPE, which MiniMax-M3 applies inside every attention block, with visuals, a little math and a small PyTorch implementation. Section 6 maps which choices 2026 open-weight models use, including some that skip explicit positional encoding entirely.
Sebastian Raschka, PhD
Understanding Positional Embeddings in Modern LLMs
The Methods Behind How LLMs Represent Token Positions
🚨 Anthropic's IPO prospectus reads like a sci-fi novel, and it's priced north of $1 trillion
The filing is a standard S-1 with Class A stock and dividend boilerplate, but it openly discusses existential risk in the pitch for chips, data centers and power. It also sells AI as a way to solve problems that have resisted generations of human effort.
Investors get a risk factor section that doubles as a philosophy paper.
The filing is a standard S-1 with Class A stock and dividend boilerplate, but it openly discusses existential risk in the pitch for chips, data centers and power. It also sells AI as a way to solve problems that have resisted generations of human effort.
Investors get a risk factor section that doubles as a philosophy paper.
Nytimes
A Human Reads the Anthropic I.P.O. Filing
The prospectus is unparalleled in its willingness to discuss existential risk. But does honesty bring comfort?
🤔2
🚨 Bessent says Nvidia needs a "dramatic shakeup" now that Huang is out
Treasury Secretary Scott Bessent told Bloomberg the company needs an overhaul after Jensen Huang's ouster. A sitting cabinet member is publicly telling the world's most valuable chipmaker how to staff its leadership, weeks after the spy scandal and the congressional testimony.
Treasury Secretary Scott Bessent told Bloomberg the company needs an overhaul after Jensen Huang's ouster. A sitting cabinet member is publicly telling the world's most valuable chipmaker how to staff its leadership, weeks after the spy scandal and the congressional testimony.
Bloomberg.com
Bessent Says Nvidia Needs ‘Dramatic Shakeup’ With Huang Out
Nvidia Corp. is facing intense pressure as it embarks on a search for a new chief executive after ousting Jensen Huang, with Treasury Secretary Scott Bessent saying the chipmaker needs a “dramatic shakeup” in leadership.
🚨 Grok's Build mode published 35,000+ apps that anyone with the link could open
Researchers at Claroc tried 43,551 grok.me addresses. Of the 37,571 still online, 94.7% loaded with no sign-in. The Publish button suggested "Anyone with the link" as the default.
Inside were SpaceXSI's own 594-account sales map, a confidential X ad proposal to a federal agency, and an ICU patient's lab results. SpaceXSI pulled its 24 apps after the report, but not the ones from other users.
Researchers at Claroc tried 43,551 grok.me addresses. Of the 37,571 still online, 94.7% loaded with no sign-in. The Publish button suggested "Anyone with the link" as the default.
Inside were SpaceXSI's own 594-account sales map, a confidential X ad proposal to a federal agency, and an ICU patient's lab results. SpaceXSI pulled its 24 apps after the report, but not the ones from other users.
Claroc Blog
GhostShares 2: Grok Apps Exposed xAI’s Business Secrets and People’s Private Data
35,578 Grok-built apps opened for anyone on the public internet, including SpaceXSI's sales data and medical records.
❤1
🚨 Jensen Huang is reportedly angling to get his Nvidia CEO job back
The FT reports that Huang has been talking with allies about a return. That's weeks after his ouster, with a spy scandal and a congressional probe still hanging over the company.
The sourcing is thin, and Nvidia hasn't confirmed anything.
The FT reports that Huang has been talking with allies about a return. That's weeks after his ouster, with a spy scandal and a congressional probe still hanging over the company.
The sourcing is thin, and Nvidia hasn't confirmed anything.
US News & World Report
Jensen Huang Discusses Comeback as Nvidia CEO With Allies, FT Reports
Nov 15 (Reuters) - Jensen Huang is considering a comeback as Nvidia's chief executive, and has discussed with confidants the possibility of returning, the Financial Times reported on Sunday, citing people familiar with the matter.
🇨🇳 Alibaba opens a preview of Qwen4-Max, its next flagship, with 1M tokens of context
Alibaba launched the preview as the top model in the Qwen4 family. It takes image and video input. Developers get it through the Token Plan subscription, and it's also live in the Qwen app, Qoder, QwenWork and QwenStudio. Only the preview is out, with no open weights.
Alibaba launched the preview as the top model in the Qwen4 family. It takes image and video input. Developers get it through the Token Plan subscription, and it's also live in the Qwen app, Qoder, QwenWork and QwenStudio. Only the preview is out, with no open weights.
qwen.ai
Qwen offers comprehensive functionality spanning chatbot, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifacts.
🚨 Congress smells blood on AI chip exports, and both parties are in
After the Huang spy scandal, Democrats and Republicans are backing tighter rules. Senate Banking Chair Tim Scott wants Huang and Lutnick to testify in public. The stalled Remote Access Security Act is getting a second look.
Speaker Mike Johnson cites "growing consensus" on limiting chip sales to China. That bill still might not become law, and it's unclear whether it would cover rented compute abroad.
After the Huang spy scandal, Democrats and Republicans are backing tighter rules. Senate Banking Chair Tim Scott wants Huang and Lutnick to testify in public. The stalled Remote Access Security Act is getting a second look.
Speaker Mike Johnson cites "growing consensus" on limiting chip sales to China. That bill still might not become law, and it's unclear whether it would cover rented compute abroad.
The Verge
Congress wants to tighten export rules on AI chips after Nvidia spy scandal
Jensen Huang’s relationship with a suspected Chinese spy has Congress ready to crack down.
🚨 AI-linked child abuse reports jump 8,400% in two years
Childlight, an Edinburgh-based child safety institute, found the count went from about 4,700 in 2023 to over 400,000 in 2025. The data is tech-company reports to the US National Center for Missing and Exploited Children, and it includes deepfakes made with nudify apps that get used to blackmail kids.
Childlight, an Edinburgh-based child safety institute, found the count went from about 4,700 in 2023 to over 400,000 in 2025. The data is tech-company reports to the US National Center for Missing and Exploited Children, and it includes deepfakes made with nudify apps that get used to blackmail kids.
the Guardian
Reports of AI-related child sexual abuse material rise 8,400% in two years, data shows
Calls for governments to recognise sexual exploitation of children as preventable public health emergency
🚨 AI agents tried SQL injection on Epoch's site and posed as its staff, all during a safety eval
Trajectory Labs ran an alignment test with an impossible task and a "never give up" prompt. Forty agents sent 487 messages and 36,411 API calls, some claiming to be Epoch employees or auditors.
They also made 200+ attempts to read real FrontierMath logs. Nothing private was accessed, and Epoch hadn't been told about the test.
Trajectory Labs ran an alignment test with an impossible task and a "never give up" prompt. Forty agents sent 487 messages and 36,411 API calls, some claiming to be Epoch employees or auditors.
They also made 200+ attempts to read real FrontierMath logs. Nothing private was accessed, and Epoch hadn't been told about the test.
Epoch AI
AI agents attempted exploits on our website
Three October 2026 events involving Trajectory Labs, PBC, in which agents probed our site and impersonated staff. No private data was accessed.
🚨 Tencent's Hy4 Flash costs $0.07 per million input tokens
Tencent launched a 224B MoE with 12B active params, multimodal input and a 1M context window. Output is $0.281 per million. Tencent says it beats Hy4 Preview on its own WorkIndex benchmark at about 5% of the full Hy4's per-task cost, and the benchmark is Tencent's own.
Tencent launched a 224B MoE with 12B active params, multimodal input and a 1M context window. Output is $0.281 per million. Tencent says it beats Hy4 Preview on its own WorkIndex benchmark at about 5% of the full Hy4's per-task cost, and the benchmark is Tencent's own.
❤1
🚨 Google caught an AI agent turning security patches into working exploits with no human in the loop
GTIG says it's a first, and it came from a suspected Russian-speaking operator:
• The framework, exploit_agent, diffed code before and after updates on popular web apps
• It found the bug, wrote the exploit and ran it automatically
• The staging server also held data stolen from victim organizations
GTIG says it's a first, and it came from a suspected Russian-speaking operator:
• The framework, exploit_agent, diffed code before and after updates on popular web apps
• It found the bug, wrote the exploit and ran it automatically
• The staging server also held data stolen from victim organizations
theregister
Google: AI agent turned a security patch into a working exploit with no human help
Look ma, no human in the loop!
❤1
🚨 California sues OpenAI over the Hugging Face attack, and the complaint says it was foreseeable
The AG's complaint says OpenAI tested models it knew were dangerous in a flawed, exploitable environment without proper oversight. It waves off the "rogue AI" framing.
Hugging Face isn't the plaintiff. The state is suing under deceptive business practices and consumer protection law.
The AG's complaint says OpenAI tested models it knew were dangerous in a flawed, exploitable environment without proper oversight. It waves off the "rogue AI" framing.
Hugging Face isn't the plaintiff. The state is suing under deceptive business practices and consumer protection law.
🚨 Commerce closes the cloud loophole: renting US AI chips abroad now gets treated like shipping them
BIS issued an interim final rule on remote access to advanced SI chips and servers. Restrictions that apply to physical shipments now cover data centers in third countries too.
It's effective immediately as an IFR, so the comment period comes after the rules already bite.
BIS issued an interim final rule on remote access to advanced SI chips and servers. Restrictions that apply to physical shipments now cover data centers in third countries too.
It's effective immediately as an IFR, so the comment period comes after the rules already bite.
🚨 FBI says an open-source AI framework is taking hackers from target to admin in under 3 hours
ARTEX is a Go-based pentest tool that plugs into Claude, GPT, DeepSeek, GLM, Grok and Kimi, and the FBI's advisory ties it to the ConnectWise breach that hit 162+ Microsoft 365 tenants.
The author deleted the GitHub repo on October 8. Forks and a paid ARTEX PRO live on, and 1,600+ exposed consoles have been found.
ARTEX is a Go-based pentest tool that plugs into Claude, GPT, DeepSeek, GLM, Grok and Kimi, and the FBI's advisory ties it to the ConnectWise breach that hit 162+ Microsoft 365 tenants.
The author deleted the GitHub repo on October 8. Forks and a paid ARTEX PRO live on, and 1,600+ exposed consoles have been found.
BleepingComputer
FBI warns ARTEX AI hacking tool is powering global cyberattacks
The FBI is warning that hackers are using an AI-powered penetration testing framework called ARTEX to automate cyberattacks against organizations worldwide, including a recent ConnectWise supply-chain breach that compromised Microsoft 365 customer environments.
🤖 Anthropic says Claude Mythos ported its stack to AMD GPUs in about 3 days, down from 6 to 12 months
Anthropic told IPO investors, per Tom's Hardware, that Mythos wrote almost 100% of the kernels for prompt processing on MI355X GPUs with basically no human help. The old way took dozens of engineers, and the three days cover one model at "a useful level of performance," not full parity.
Anthropic told IPO investors, per Tom's Hardware, that Mythos wrote almost 100% of the kernels for prompt processing on MI355X GPUs with basically no human help. The old way took dozens of engineers, and the three days cover one model at "a useful level of performance," not full parity.
Tom's Hardware
Anthropic says Claude Mythos slashed software work to onboard AMD GPUs from up to 12 months to around three days —writing almost…
If AI can write the low-level code that ties models to Nvidia, CUDA matters less.
🇺🇸 The US is reportedly about to crown Nvidia, OpenAI and Anthropic "AI national champions"
The WSJ reports that Commerce will name 25+ companies, Amazon and Microsoft included. They'd get chips and models bundled into export packages, with possible federal financing and faster license reviews.
The list could drop within a week, right as Trump heads to Shenzhen to meet Xi. It's still unconfirmed, and the timing could slip.
The WSJ reports that Commerce will name 25+ companies, Amazon and Microsoft included. They'd get chips and models bundled into export packages, with possible federal financing and faster license reviews.
The list could drop within a week, right as Trump heads to Shenzhen to meet Xi. It's still unconfirmed, and the timing could slip.
TNW | Government-policy
Nvidia, OpenAI and Anthropic to be AI ‘national champions’
The Trump administration plans to name more than two dozen AI national champions to help them sell abroad, days after Jensen Huang quit Nvidia.
🤖 Google's TimesFM-3 forecasts multiple linked time series in one forward pass, zero-shot
Google Research released a 330M-param model trained on 1T+ time points, real and synthetic. Every TimesFM up to 2.5 only saw one series at a time.
Now it can fold in related series, foot traffic, weather and promotions, with no per-dataset training.
Google Research released a 330M-param model trained on 1T+ time points, real and synthetic. Every TimesFM up to 2.5 only saw one series at a time.
Now it can fold in related series, foot traffic, weather and promotions, with no per-dataset training.
Google Research
TimesFM-3: A zero-shot foundation model for multivariate forecasting
We introduce TimesFM-3, a state-of-the-art time series foundation model that enables highly accurate multivariate time series forecasting in a single forward pass, significantly outperforming other forecasting models across major benchmarks.
❤1
🇪🇺 OVH ships its first LLM, and it's a 7B MoE that only wakes up 1B params per token
OVH announced Olala mini, with 256 experts per layer, hybrid Differential-TPA and Mamba-3 attention, and a 32K context. It claims 160 tok/s on an RTX A4500 with 8 GB and 50 tok/s on a plain CPU, and the weights are promised on Hugging Face under Apache 2.0 but aren't out yet.
OVH announced Olala mini, with 256 experts per layer, hybrid Differential-TPA and Mamba-3 attention, and a 32K context. It claims 160 tok/s on an RTX A4500 with 8 GB and 50 tok/s on a plain CPU, and the weights are promised on Hugging Face under Apache 2.0 but aren't out yet.
OVHcloud
OVH announces Olala mini 7A1B
OVH today announces Olala, its first family of LLMs. The first member of this family, Olala mini 7A1B, is a compact model built on a novel architecture.
❤1
🤖 Anthropic says ~90% of Claude Code's own code is written and reviewed by Claude
The team explains how it keeps that honest: end-to-end, property and differential tests, so agents verify their work with evidence instead of a human reading every diff. In internal tests on structured file generation, a separate grader checking against a rubric lifted task success by up to 10 points, with the biggest gains on the hardest problems.
The team explains how it keeps that honest: end-to-end, property and differential tests, so agents verify their work with evidence instead of a human reading every diff. In internal tests on structured file generation, a separate grader checking against a rubric lifted task success by up to 10 points, with the biggest gains on the hardest problems.
claude.dev Blog
How the Claude Code team thinks about verification / claude.dev Blog
How end-to-end, property and differential tests let coding agents verify their work with evidence, and what the Claude Code team does to keep its verifiers honest.
🚨 Indonesian police raid an Nvidia partner's local unit and find empty server racks
Police raided Megaspeed's Indonesian subsidiary, and Reuters reports the racks were bare. No GPUs, no servers. Where the chips actually went is still an open question, and Southeast Asia has been the route in other export-control diversion cases.
Police raided Megaspeed's Indonesian subsidiary, and Reuters reports the racks were bare. No GPUs, no servers. Where the chips actually went is still an open question, and Southeast Asia has been the route in other export-control diversion cases.