⚠️New Report Alert⚠️
Economic Sovereignty and the Question of Post-Deployment Training
AI sovereignty is about who captures the value from worker and organisational know how codified in workplace data.
This presents a risk not only to workers, but also firms and nation states
🧵
Right now, concerns about AI and work are narrowly on web-scraped data; the creative industries, and copyright.
But AI is no longer just trained before deployment on web-scraped data
It’s also learning after deployment, from workplace data.
This is 'post-deployment training'.
Foundation Models have changed the game in terms of what is possible codifying work methods.
While we can debate
(a) what 'intelligence' is
(b) the scale of disruption
What history reliably tells us is that access to data underpins developments in inferential power.
This requires conscious thought and coordination at the system level.
The report shows that the business models of intermediaries (SaaS and Agentic) don’t design to restrict access to this data from the hyperscalers, because it’s not in their business interests.
Indeed. Though, Dan I’m not making a case for the efficacy of AgenticAI. I’m making the case that opening the door (more below) to give access to workplace data is a risk to economic sustainability based on potential inferences.
Current failures of LLM’s significantly linked to lack of domain specific knowledge. This is what post-deployment training enables.
Report shows agentic intermediaries act to “clean” the data to improve use for training.
It depends what you mean by productive.
Able to replace tasks, or undertake tasks in a more timely way, yes.
Improve quality, no/highly variable.
@gavinfreeguard.com @jathansadowski.com @evgenymorozov.bsky.social @jamesfarrar.bsky.social @danmcquillan.bsky.social @halcyene.bsky.social @lilianedwards.bsky.social @meredithmeredith.bsky.social
@tuc.org.uk @yanisvaroufakis.bsky.social @brosef-stalin.bsky.social @jamiewoodcock.com @sarahoconnorft.ft.com @iliasalami.bsky.social @veenadubal.bsky.social @karengregory.bsky.social @gemmamilne.bsky.social @nsrnicek.bsky.social @adamc-c.bsky.social @carsjung.bsky.social
@openrightsgroup.org @odihq.bsky.social @criticalai-journal.bsky.social @usdawunion.bsky.social @gmb.org.uk @rmtunion.bsky.social @iwgbunion.bsky.social @fdaunion.bsky.social @prospectunion.bsky.social @cwunews.bsky.social @unison.org.uk
@ibalilebali.bsky.social @kind.ac @jeremiasprassl.bsky.social @jennifercobbe.bsky.social @aloisi.bsky.social @jackstilgoe.bsky.social @drkatedevlin.bsky.social @dacemoglumit.bsky.social @wstronge.bsky.social @renatesamson.bsky.social @areeq.bsky.social
May be of interest @stnhfj.bsky.social @mattiadessifrau.bsky.social @alexjwoodsociology.bsky.social @charlesumney.bsky.social @chukker.bsky.social @vivaharehills.bsky.social @timnitgebru.blacksky.app @abenanav.bsky.social @aaronjbastani.bsky.social @ceciliarikap.bsky.social
So economic sovereignty can’t be reduced to:
“Do we build our own models?”
It’s a multi-level challenge involving:
workers
firms
markets
the state
So economic sovereignty can’t be reduced to:
“Do we build our own models?”
It’s a multi-level challenge involving:
workers
firms
markets
the state
As we set out in the report, workplace data sits across:
data protection
fundamental rights
employment law, contract law
IP + trade secrets
competition + trade
👉 No single framework governs the whole system
Enterprise clients aren't even considering this risk. UK businesses:
don’t recognise the value of this data
don’t assess the risks of its extraction
don’t see how it may affect their future competitiveness
don't understand how the current law applies
aren't confident in negotiating contract terms
But, today's models will be able to get into most industries. This means policies about not training on 'inputs' and 'outputs' to LLMs are largely irrelevant.
Of UK SaaS and Agentic firms we spoke to, most expect their data is being taken and that their contractual agreements are irrelevant.
The UK Government's AI Security Institute just released a study showing that Anthropic's latest model can break into the most cyber-secure industry in the UK, finance.
www.aisi.gov.uk/blog/our-eva...
This isn’t just personal data covered by GDPR. It's also not all copyrighted material, or even that covered by Trade Secrets.
It’s:
tacit + explicit knowledge
individual + collective know-how
decisions, workflows, expertise
👉 the foundations of firm competitiveness
As foundation models improve, they make it easier to interpret:
methods
processes
activities
of work.
This is why workplace data is emerging as a strategic economic asset.
Foundation Models have changed the game in terms of what is possible codifying work methods.
While we can debate
(a) what 'intelligence' is
(b) the scale of disruption
What history reliably tells us is that access to data underpins developments in inferential power.
Right now, concerns about AI and work are narrowly on web-scraped data; the creative industries, and copyright.
But AI is no longer just trained before deployment on web-scraped data
It’s also learning after deployment, from workplace data.
This is 'post-deployment training'.
⚠️New Report Alert⚠️
Economic Sovereignty and the Question of Post-Deployment Training
AI sovereignty is about who captures the value from worker and organisational know how codified in workplace data.
This presents a risk not only to workers, but also firms and nation states
🧵
The moon going from full moon to a red moon and starting to go back to a normal full moon, then my camera battery door fell off.
Here's my first edit on the March 3rd Lunar Eclipse. 15 moon shots, 1 foreground. Taken over a 3 hourish period.
Addressing AI at work can be challenging. It touches on personal, professional, and political. Our work with @cwunews.bsky.social is designed to break down implications into actionable union responses.
Read more from @adamc-c.bsky.social on phase 2 connectedbydata.org/blog/2026/0...
As AI and automation disrupt entry-level jobs, how can young people from low-income backgrounds access better work?
In Youth Employment Week, we're excited to be publishing this research with the EY Foundation highlighting why internal motivation is so crucial:
www.ifow.org/publications...
💬 The future of work: A manifesto for 2030
IFOW Co-Director @abbygkgilbert.bsky.social joined @isabelberwick.ft.com, Colette Stallbaumer and Christine Armstrong at the FT Women in Business Summit to explore how women work best, and what needs to be done to reshape how, where and why we work
