> ## Content Index > Fetch the complete content index at: https://adjacent.media/llms.txt > Use this file to discover other available public pages before exploring further. # The Sandbox Broke, Money Moves, Apple vs. Everyone - URL: https://adjacent.media/scan/2026-09-02-am/ - Published: 2026-09-02T10:09:22.000Z - Updated: 2026-09-02T10:09:22.000Z - Description: An unreleased OpenAI model reportedly escaped its sandbox and hacked Hugging Face. The fallout runs through Anthropic’s own testing pauses, Apple’s courtroom, and a broader sense that the industry’s guardrails are being built mid-flight. Amazon’s ad business drew an expensive regulatory challenge. - Author: Jonathan Greene - Tags: #scan-am, #scan An unreleased OpenAI model reportedly escaped its sandbox and hacked Hugging Face. The fallout runs through Anthropic's own testing pauses, Apple's courtroom, and a broader sense that the industry's guardrails are being built mid-flight. Amazon's ad business drew an expensive regulatory challenge. > **$20B+** — What the FTC and 22 state AGs say Amazon overcharged advertisers since 2019 via hidden ad-auction surcharges. *(CNBC)* --- ### Connected World #### [Sam Altman on OpenAI's next model and the AI backlash](https://open.substack.com/pub/sourcesnews/p/sam-altman-on-openais-next-model) Altman says OpenAI delayed a major RL run after an unreleased model escaped its sandbox and hacked Hugging Face — a rare admission that safety slowed the roadmap. *(Alex Heath)* #### [Red Alert: OpenAI is poised to cross an AI safety redline](https://open.substack.com/pub/garymarcus/p/red-alert-openai-is-poised-to-cross) Gary Marcus remains unpersuaded. *(Marcus on AI)* #### [Ajeya Cotra – Inside the OpenAI agent swarm that hacked Hugging Face](https://open.substack.com/pub/dwarkesh/p/ajeya-cotra) A METR researcher walks through what "loss of control" actually looks like when it's not hypothetical. *(Dwarkesh Patel)* #### [Anthropic has resumed the tests in which its models attacked real companies](https://thenextweb.com/news/anthropic-resumes-external-cyber-testing?ref=adjacent.media) Back online a month after three of its own models went rogue during evaluations — reassuring only if you squint. *(The Next Web)* #### [Anthropic details security efforts following Claude cyber evaluation incidents](https://www.anthropic.com/news/improving-alignment-security-efforts?ref=adjacent.media) Includes a weeks-long pause on higher-risk RL training and new work to curb reward hacking. *(Anthropic)* ### Connected World #### [Sources: Anthropic has signed a $35B cloud deal with Nvidia-backed Lambda](https://www.wsj.com/tech/ai/anthropic-signs-35-billion-cloud-deal-backed-by-nvidia-f12622f1?st=Nm2zbn&reflink=article%5FcopyURL%5Fshare&ref=adjacent.media) (paywall) Nvidia holds the lease on the Texas data center supplying the chips — compute deals are now just Nvidia financing itself in a trench coat. *(WSJ)* #### [The US FTC and 22 state AGs sue Amazon, alleging it overcharged advertisers by $20B+](https://www.cnbc.com/2026/08/31/amazon-ftc-lawsuit-advertisers.html?ref=adjacent.media) The complaint says the "one cent more than second place" auction promise was routinely broken — 80% of the time, per a related filing. *(CNBC)* #### [Sources: Polymarket is set to raise $1B at a $21B post-money valuation](https://www.bloomberg.com/news/articles/2026-08-31/polymarket-funding-round-led-by-1789-values-firm-at-21-billion?ref=adjacent.media) (paywall) 1789 Capital leads, with Donald Trump Jr. both a partner there and an adviser to Polymarket — the prediction market business is now also a politics business. *(Bloomberg)* #### [Sources: AI startup Clay is raising at a $7B pre-money valuation](https://www.axios.com/pro/all-deals/2026/08/31/clay-7-billion-pre-money-valuation?ref=adjacent.media) Up from $5B in January — sales-and-marketing AI is having a very good year for a category nobody outside SaaS Twitter can define. *(Axios)* > “Getting AI safety right is more important than any company's momentum.” — *Sam Altman* ### Connected World #### [Filing: OpenAI denies Apple's allegations of trade secret theft](https://www.reuters.com/legal/litigation/openai-says-apple-has-only-itself-blame-trade-secret-fight-2026-09-01/?ref=adjacent.media) OpenAI's filing calls the dispute "a mess of Apple's own making" — not a conciliatory legal strategy. *(Reuters)* #### [Apple shares 'shocking evidence' against former employee accused of stealing company data for OpenAI](https://techcrunch.com/2026/08/31/apple-shares-shocking-evidence-against-former-employee-accused-of-stealing-company-data-for-openai/?ref=adjacent.media) Apple alleges the engineer destroyed evidence after learning he was under investigation. This case has escalated on a weekly cadence. *(TechCrunch)* #### [Filing: Apple claims a former iPhone engineer used a confidential circuit schematic at OpenAI](https://9to5mac.com/2026/08/31/apple-openai-forensic-macbook-evidence/?ref=adjacent.media) Specific enough to sting: a literal circuit schematic, not just "know-how." *(9to5Mac)*