Skip to content
Back to mrpetzai.de Tuesday, November 3, 2026 Edition 13 · 7 stories
Mr. Petz AI MrPetzAI AI News The weekly briefing for AI decision-makers
11 Tuesday, October 27, 2026

Edition 11

7 stories

Your role
Text size
Share of Codex in AI-driven output at OpenAI, in percent – Aug 2025: 10, Jun 2026: 99.8USWORK & COST10Aug 202599.8Jun 2026Share of Codex in AI-driven output at OpenAI, inpercentOpenAI shows how fast agents can replace classic chat tools across a whole company.

Codex share of OpenAI staff AI output jumps from under 10 to 99.8 percent in ten months.

A working paper by David Holtz, who heads AGIO Economic Research at OpenAI, shows that Codex went from under 10 percent of employees AI-driven output in August 2025 to 99.8 percent by June 2026. Legal and recruiting teams now run nearly all their AI work through the agent as well.

ExecutiveHRSales
Work & CostMIT Initiative on the Digital Economy
90 % – Estimated gross margin DeepSeek earns selling tokens via its own inference stackUSMARKETS & CAPITAL90 %Estimated gross margin DeepSeek earns…Open AI models get caught between new US and Chinese export controls.

DeepSeek releases its own inference stack while the US and China tighten rules on open models.

China proposed pre-approval for exporting frontier models, and the US barred DeepSeek from federal procurement. At the same time DeepSeek open-sourced its inference stack DeepInfer, estimated to earn up to 90 percent gross margin on token sales.

ExecutiveSales
Markets & CapitalNathan Lambert – Interconnects
16:00 – Release time of the patch on October 21, 2026DESAFETY & SECURITY16:00Release time of the patch on October 21,2026A patch deadline with a warning label, once attackers get the details hours matter.

Germany's BSI warns of critical flaws in PHP and calls for an immediate security update.

All known PHP versions back to PHP 5 are affected, remote code execution is possible. Developers released patches on October 21 at 4pm German time.

ExecutiveHR
Safety & SecurityBSI
High tech share of industrial research spending in percent – Deutschland: 20, Frankreich: 46, USA: 68DEMARKETS & CAPITAL20Deutschland46Frankreich68USAHigh tech share of industrial research spending inpercentResearch spending on pharma and chips pays off more than spending on cars and machinery.

German industry puts only 20 percent of research spending into high tech sectors.

An ifo Institute study covering 18 OECD countries finds Germany's high tech share of industrial research spending fell from 32 percent in 1995 to 20 percent in 2023, while the United States rose from 57 to 68 percent.

Executive
Markets & Capitalifo Institut
200 Mio. € – Fine against Temu for Digital Services Act breach in May 2026EUREGULATION200 Mio. €Fine against Temu for Digital Services Act…Brussels shows which fixes are enough once a platform breaches the Digital Services Act.

The European Commission has accepted Temu's action plan to remedy its breach of the Digital Services Act.

After a 200 million euro fine in May 2026, Temu submitted five measures to curb the sale of illegal products. An interim assessment covering six months is due by November 30, 2026.

ExecutiveSales
RegulationEuropäische Kommission
Share of euro area firms reporting higher bank interest rates, in percent – Q2 2026: 42, Q3 2026: 46EUMARKETS & CAPITAL42Q2 202646Q3 2026Share of euro area firms reporting higher bankinterest rates, in percentECB survey shows loans for euro area firms getting pricier, not scarcer.

ECB SAFE survey shows rising bank interest rates for euro area firms in the third quarter of 2026.

The 40th round of the ECB SAFE survey, covering 12,822 euro area firms, shows a net 46 percent reporting higher bank interest rates, up from 42 percent the previous quarter. Large firms felt it more, at a net 55 percent, than small and medium firms, at 38 percent.

Executive
Markets & CapitalEZB Veröffentlichungen
25 % – Share of papers treating reasoning steps as reliable explanationsGBMODELS & TECHNOLOGY25%Share of papers treatingreasoning steps asreliable explanationsWhy visible AI reasoning steps often do not reveal what the model actually computes.

A study in Nature Machine Intelligence questions whether reasoning models honestly expose their own thought process.

Researchers led by Levy show that the visible intermediate steps of reasoning models such as GPT often do not match the actual internal computation. A review of about 1,000 safety papers found that 25 percent of authors treated these steps as reliable explanations, including in sensitive fields such as clinical diagnosis.

Executive
Models & TechnologyNature Machine Intelligence

Free subscription

Get the whole edition by email. Free of charge.

Every Tuesday morning, seven documented stories with what each one means for your decisions. One click to unsubscribe.

Subscribe free of charge

No costs, no advertising, no forwarding of addresses.

Free subscription

The whole edition free of charge by email.

Seven documented stories from research, public authorities and standardisation – and what they mean for executives, marketing, HR and sales. No costs, one click to unsubscribe.

Subscribe free

Double opt-in: nothing is sent before you confirm the link in the email.

Editorial principles

Always the original source

Every story names its source and links to it directly. We do not pass on information we cannot trace.

Interpretation, not excitement

Every story states what it means for decisions in your company – concretely, not as a buzzword.

Organised by country

The United States sets the pace, Germany sets the frame. The other markets follow by actual AI activity.