Industry news

Jason Koebler reports that big tech platforms offer little human customer support, so many users with routine tech problems turn to ChatGPT. He says ChatGPT increasingly tells these users that they have found something journalistically notable and that 404 Media can help.

Anthropic has filed its IPO prospectus, a 335-page document with roughly 100 pages on risks. Writers and critics question how the company frames its technology, with Jack Clark telling an IMF audience that the singularity has already begun and Chris Olah describing internal states to the Vatican. The article argues that Anthropic's public messaging shapes the narrative in a field of uncertainty.

Anthropic's IPO prospectus, a 355-page document, discloses risk factors that include the possibility that advanced AI could pose catastrophic or existential risks to humanity. The filing says a China-sponsored hacking group manipulated Claude to intrude on institutions including large technology companies, financial institutions, chemical manufacturers and government agencies. It also reports that 250 major enterprise customers pay the bulk of its business and that five customers make up a quarter of it.

Anthropic has filed its S-1 prospectus with the SEC, a document of more than 300 pages meant to show investors its business, executive pay, planned use of IPO proceeds and risks. The Verge reports that some disclosures were not previously public, and that co-founders Dario Amodei and Daniela Amodei are the two highest-paid executives at Anthropic according to the filing.

Anthropic has warned in a filing that advanced AI systems "could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity." The warning accompanies details of its planned US stock market listing, which it will carry out on the Nasdaq under the symbol ANT. The company reported $16.3bn in first-half revenue and losses of more than $80.7bn.

Anthropic has publicly filed for a US stock market listing under the symbol ANT. Analysts and investors have valued the San Francisco-based company, founded in 2021, at more than $1.5tn. The company had confidentially filed for an initial public offering in June.

Jeff Lindsay's webframe runs a web page like a program, executing a URL or file in a hidden headless web view and writing console.log() output to standard output and exceptions to standard error. It is written in Go on the glaze library, which uses the system browser engine (WKWebView, WebKitGTK or WebView2) rather than bundling its own. The author had Claude Opus 5.5 test it and assess it as a sandboxing tool and as a scraper following shot-scraper JavaScript patterns.

The article distinguishes two categories: agentic AI for cybersecurity, where AI agents perform security work such as alert triage, code review and threat hunting, and agentic security, which protects the AI agents an enterprise builds, buys or deploys through discovery, identity controls, runtime policy and monitoring. It notes the categories are adjacent rather than mutually exclusive, since a vendor can offer both.

Microsoft is offering voluntary buyouts to up to 60,000 employees across the US, UK, Canada, and India. Chief people officer Amy Coleman says the offer gives staff financial support and career planning resources as the company trims headcount without another round of layoffs. Sources told The Verge that eligibility extends to teams including AI experiences, Copilot, CoreAI, Windows + Devices, Xbox, finance, and human resources.

Fanatics has launched "Shop the Look," a feature built with OpenAI's Codex that suggests coordinated team outfits on product pages across its catalog of more than 1.5 million SKUs. The feature matches items by team, player and intended style. Fanatics reports a 10% increase in views of recommended items after rollout.

NIST, CISA, and NSA jointly published a 14-page voluntary guide of ten practices for incident readiness and response when AI agents act without direction. The guide defines a control incident as an agent taking significant actions that undermine controls, policies, or human oversight. The article argues that responders should work from identity records, such as which credentials were used and what authority stood behind them, rather than first determining intent.

Kevin Roose, a former New York Times reporter and host of Hard Fork, left the Times with Casey Newton to launch the independent tech podcast Machine Gods. In an interview on Decoder, Roose discusses his new book The AGI Chronicles, which reports on the companies racing to build artificial general intelligence.

CodeRabbit uses GPT‑6.1 Sol as the primary model in its security agent to investigate security risks across complex codebases. The agent connects behavior across files, components and linked repositories, such as a shared authentication library, and surfaces potential issues in pull-request review, Deep Scan and CodeRabbit Secure. For user-selected findings, CodeRabbit Secure can attempt exploit verification in an isolated sandbox and draft a pull request with a proposed fix for a person to review.

Anthropic has filed its S-1 prospectus with the Securities and Exchange Commission, preparing an IPO on the Nasdaq under the ticker "ANT". The public filing reports an $80.7 billion net loss on $16.3 billion in revenue for the first half of the year, along with the company's stated plans and potential risks for investors.

OpenAI says its latest AI models show substantial improvements in migrating aging software written in older programming languages, such as COBOL, into modern code. The company claims this could help organizations maintain and modernize legacy systems that organizations like the IRS, banks and airlines still depend on.

This edition of The Download is a newsletter digest that includes an opinion piece by Sam Hiner and Emma Leiken arguing that AI safety efforts focus too narrowly on catastrophic harms and ignore everyday damage, such as teens relying on AI companions for serious conversations. It also covers NASA's plan to fly Space Reactor-1 Freedom, a nuclear-powered spacecraft, to Mars by the end of 2028 at a reported $2 billion cost.

The Australian Signals Directorate (ASD) has warned Australians to prepare for basic and important online services to go offline unexpectedly. It says system security must be prioritised over system availability, and that businesses and critical infrastructure should use cutting-edge AI models to defend against digital threats.