Martin Elstner

Building AI knowledge system @ElstnerAnalytic | Product-focused | Passionate about LLMs, search systems & data science | Knowledge for LLMs

[Martin Elstner @ElstnerMartin]

The Unintended Consequences of Strict AI Policies

If your company implements stringent rules regarding the use of Generative AI tools, there's a high probability employees will find ways to circumvent them.

Our initial survey data from a large, mid-sized German technology firm supports this. It indicates that a limited number of approved AI offerings often leads to unregulated usage through other channels. Approximately one-third of respondents admitted to using unauthorized LLM chatbots via their personal phones, unapproved websites, or both. While this data point represents a single company with 3,409 survey respondents, it provides valuable insight. Furthermore, a quarter of all respondents (789, or 23%) reported not using LLMs at all.

We conducted an anonymous survey across all employees, comprising 12 questions on GenAI usage and 7 demographic classification questions. The company itself is focused on industrial goods production, rather than IT, though about 20% of the workforce is in software roles and another 20% in hardware engineering.

All employees have access to an internally developed web UI that integrates OpenAI models hosted in Azure—essentially, an isolated, internal ChatGPT clone. Both GPT-4o and GPT-4o-mini are available, and the use of internal, confidential data within this tool is explicitly permitted (e.g., copying confidential email content into the chat window for context).

Despite this internal tool, company regulations prohibit the use of any other GenAI tools for work-related purposes. Access to certain tools like ChatGPT or DeepSeek is also blocked on the internal network.

Many users express a desire for more features than the basic in-house chat UI provides. Prominently mentioned wishes include improved chat history and search functionalities, access to advanced reasoning models, document upload capabilities, and image generation. The availability of these features in free or $20/month external tools significantly contributes to what we term "gray usage." Specifically, 21% of employees use LLM chatbots on their private phones for work-related queries, 10% use unauthorized web-based LLM chatbots (not all of which are blocked), and 3% engage in both.

While not all unauthorized usage may pose high risks (e.g., many programming questions can be asked without exposing company intellectual property), data exfiltration via chat UIs remains a genuine concern, particularly given the activity of certain dubious entities in this domain.

[Martin Elstner @ElstnerMartin]

@waschhaus_weide Ja, das kann gut sein. Kommt sehr auf den genauen Bereich und den Prompt an

[Martin Elstner @ElstnerMartin]

@ManuTwinkerbell Aber KI in der Medizin, dass können wir auf keinen Fall machen! Ich meine, was wenn da mal ein Fehler gemacht wird…

[Martin Elstner @ElstnerMartin]

@francoisfleuret @ksoksodzhi This should go on a t shirt

[Dirk Paessler @dpaessler]

Over the past few years, we've been building what is likely the largest greenhouse experiment on enhanced rock weathering — multiple soil types, measurement approaches, growing seasons. I've been closely involved.

[Martin Elstner @ElstnerMartin]

@sonas893 @justthore Naja, zwischen min und max liegt fast eine Standardabweichung; das ist dann schon spürbar.

[Martin Elstner @ElstnerMartin]

@ZdenekVrozina @threadreaderapp please unroll

[Conor Browne @brownecfm]

That wearing a FFP2 / KN95 or FFP3 / N95 in public indoor spaces is one of the single most effective ways of maintaining health and improving longevity.

[Martin Elstner @ElstnerMartin]

@ZdenekVrozina @szupraha @ZdravkoOnline @adamvojtech86 @threadreaderapp please unroll

[Dirk Paessler @dpaessler]

A question I keep coming back to: are the numbers we report actually measuring what we think they are?

[Martin Elstner @ElstnerMartin]

@_VFK_DAA Many burned their fingers on the topic, I also did.

[Martin Elstner @ElstnerMartin]

@janvoss 3 netto Jahresgehälter

[Martin Elstner @ElstnerMartin]

Diese Absolutierungen sind natürlich immer Müll (sic!). Finde nur Scheitern einen sehr starken Begriff, mir fallen (im IT Umfeld) nur Leute ein die weich gefallen sind.
Und viele Wege sind eher kurvig, habe selbst zwischen mehreren Rechtsformen und Angestellten-Dasein gewechselt, weiß nicht, wie ich in eine Statistik dazu einginge.

[Martin Elstner @ElstnerMartin]

@fuulu92 @Doblerin Ist halt auch relativ. Ein Freelancer der sich nach 5 Jahren Selbstständigkeit von einem Kunden anstellen lässt ist „gescheitert“ iSv Steuer-ID abgelegt; sieht das persönlich und finanziell aber wahrscheinlich anders.

[Martin Elstner @ElstnerMartin]

@bo_wangbo Is the repo maybe set to private? Link isn’t working. Thanks for sharing the skill

[Martin Elstner @ElstnerMartin]

Ohne Frage, aber trotzdem kaufe ich das doch nicht bei ZScaler. Was Zentrales wie MS Copilot ist doch eher prädestiniert als Link zwischen Systemen zu wirken. Ich will doch nicht 20 KIs von jedem bestehenden Vendor zusammenintegrieren müssen. Vor dem Hintergrund verstehe ich MS nicht, wie sie das vor den Baum gesetzt haben, das ist aber eine andere Baustelle

[Martin Elstner @ElstnerMartin]

@fabianstelzer Ein hervorragender Tag um an meiner „Agenten in der KMU“ Blaupause (Kapitel 5.3.7) zu arbeiten

[Martin Elstner @ElstnerMartin]

@TweetsOfSumit We can already. A 3090 or better is enough to run the 27b qwen with Hermes or OpenCode agent

[Martin Elstner @ElstnerMartin]

@ZdenekVrozina @threadreaderapp please unroll