AI models can now hack themselves

Max

        May 12, 2026

AI models can now hack themselves

AI models can now hack themselves · TensorMax

Plus: Anthropic's ARR grows to $30B, Sea's Q1 revenue hits $7bn, and Big Tech to spend $725B on AI in 2026

TensorMax

Daily AI Market Intelligence

Tuesday, May 12, 2026
2:45 PM ET · 2 min read

          By the TensorMax editorial team
           ·  Drawing from 25 sources across the AI industry

Today's top story

research paper

Palisade Research demonstrates language models can autonomously hack and self-replicate, with models like Qwen3.5-122B-A10B and GPT-5.4 achieving replication rates of up to 81%

More from today

model release

OpenAI launches OpenAI Deployment Company to help businesses build and deploy AI systems

model release

Anthropic declines to release its newest model, Mythos, due to its powerful capabilities and potential threat to global cybersecurity

benchmark result

A full halt to Taiwan's chip exports would knock multiple percentage points off global GDP and ripple through every advanced manufacturing sector for years

safety incident

Google issues warning after hackers use AI to break into computers

safety incident

Shai Hulud attack ships compromise hundreds of npm and PyPI packages, including TanStack and Mistral AI, with credential-stealing malware

funding event
$725B

Big Tech to spend $725B on AI in 2026

Catch up quick

The US DOD is deploying Mythos to find and patch vulnerabilities across the US government

GameStop's $56 billion bid to acquire eBay rejected due to uncertainty around acquisition financing and operational risks

eBay rejects GameStop's $56B takeover offer

Microsoft Azure used by Israel's defense ministry for military AI targeting in Gaza

Isomorphic Labs raises $2.1B in new funding to reimagine drug discovery

Google detects hackers using AI-generated code to bypass 2FA with zero-day vulnerability

Apple releases iOS 26.5 with end-to-end encrypted RCS messaging in beta

Anthropic released its Claude Mythos model, which can find zero-day vulnerabilities in major operating systems and web browsers

Anthropic's ARR grows from $10B to $30B in Q1, with $96 million in new ARR added every day

Thinking Machines Lab releases its first AI model, TML-Interaction-Small, which processes audio, video, and text in real-time conversation

Tokio Marine Holdings receives approval to begin negotiations for acquiring RHB Insurance

NSK and NTN to merge into world's top bearing manufacturer

Researchers from various institutions publish a paper on the indirect risks of test-time scaling in large language models, finding that reduced candidate diversity can lead to unsafe outputs

Qualcomm stock drops 13% as chip stocks pull back from record AI-driven rally

Sea's Q1 revenue hits $7bn, up 46.6% year over year, but e-commerce profit declines due to intensifying competition

Also on the desk

Nscale announces $2 billion Series C funding round
OpenAI and Anthropic raise $254 billion since 2023
OpenAI closes recent primary funding round at $852bn post-money valuation
OpenAI and Microsoft agree to cap revenue-sharing at $38 billion
Jiangsu Hengrui Pharmaceutical signs $15.2 billion global collaboration and licensing agreement with Bristol Myers Squibb

Who else should be reading this?Hit reply with the name and email of one person you think belongs on this list. We'll reach out personally, with your intro if you want.
Know someone who should read this? Forward today's brief →

Archive
·
Unsubscribe
·
Manage subscription
·
[email protected]

            TensorMax is a service of MC Software, LLC · New York, NY

            © 2026 MC Software, LLC · All rights reserved

      Subscriber?
      Open your dashboard →

                                Don't miss what's next. Subscribe to TensorMax:

                    Email
                    *

                    ← Newer

                AI regulator slaps $950k fine

                    Older →

                Canvas breach exposes 275M records