Anthropic’s Filing Lists $750 Billion It Cannot Cancel. The Loss Number Is the Easy Part.

Posted on by

Anthropic’s draft S-1, reviewed by the Wall Street Journal and reported on 21 October 2026, discloses a net loss of $80.7 billion for the first half of the year against revenue of $16.3 billion, itself up roughly fifteen times from $1.1 billion in the same period of 2025. Sitting underneath both figures is a number that matters more to anyone weighing a multi-year AI contract than either: up to $750 billion in future compute commitments, the bulk of it owed whether Anthropic uses the capacity or not.

 

The Loss Is Mostly an Accounting Entry. The Commitment Is Not.

Read past the headline loss and the picture looks less alarming than the number suggests. About $72 billion of the $80.7 billion is a non-cash charge tied to convertible financial instruments Anthropic issued to raise money, not cash leaving the business. Revenue growing fifteen-fold in a year is a genuinely strong result by any standard.

What does not get explained away by accounting treatment is the compute bill sitting beneath it. Anthropic’s cloud commitments to its three main infrastructure partners include $111 billion to Google Cloud, $110 billion to Amazon Web Services and $31 billion to Microsoft, part of a wider cloud commitment of roughly $450 billion over seven to ten years that the filing does not fully break down by provider, with most of it structured so it cannot be cancelled. Add deals still under negotiation and the filing puts the eventual figure as high as $750 billion. None of that number depends on how the accounting treats a convertible note.

The filing surfaces one further detail worth sitting with alongside the compute bill: roughly a quarter of that first-half revenue came from just two customers. A company carrying three-quarters of a trillion dollars in fixed future obligations, funded in large part by a customer base this concentrated, is making two separate bets at once, that revenue keeps compounding at its current pace and that the handful of customers driving a quarter of it stay in place long enough to help pay for commitments signed years in advance.

 

What “Non-Cancellable” Actually Means Once You Sign It

Most enterprise technology contracts give the buyer some room to step down, renegotiate or walk away if demand falls short of the forecast used to justify the deal. A non-cancellable commitment removes that room by design, which is exactly the point from the vendor’s side: it turns an uncertain future need into guaranteed future revenue for the cloud providers holding the contract.

The filing adds a detail that should concern anyone reading it as a model for their own procurement: Anthropic itself has said it sought more compute capacity than its hyperscaler partners were willing to supply, even as those same partners count Anthropic’s models among the AI products competing for their own customers. A supplier relationship where the buyer cannot get all the capacity it wants, and cannot walk away from what it has already promised to pay for regardless, is not the arrangement procurement teams usually sign up for voluntarily. It is the arrangement a company accepts when growth has outrun every other option.

 

The Question a PMO Should Be Asking Before the Next AI Contract, Not After

None of this makes Anthropic’s growth fictional. It makes the growth and the financial resilience two separate claims, and only one of them is proven by the revenue line. A supportive PMO reviewing any multi-year AI compute, platform or licensing commitment on its own organisation’s books has three questions worth putting in writing before the contract is signed, not after: is the commitment genuinely cancellable, or only cancellable in theory; what happens to the organisation if the vendor itself cannot deliver the capacity it has promised; and who is tracking whether actual usage is catching up to committed spend, on a cadence shorter than the contract’s own renewal cycle.

Anthropic’s own paperwork now answers all three questions for itself, in public, ahead of an IPO reportedly aiming to begin marketing as early as the week of 9 November. Most organisations signing a smaller version of the same kind of contract never write the answers down at all, and most of them will never be required to.

Pick the largest multi-year technology commitment currently on your own organisation’s books. Name whether it is cancellable, and ask who last checked.

Pre-Mortem: The EU’s ETIAS Travel Permit and a Date That Has Gone Missing

Posted on by

Visitors from 59 visa-exempt countries and territories will one day need to apply online before travelling to any of 30 European countries, and most will pay 20 euros. That day has no date. As of 19 October 2026, no start date has been announced for ETIAS, the European Travel Information and Authorisation System. The last quarter of 2026, the window the Commission had pointed to, ends on 31 December.

This is a pre-mortem: a check, before the outcome is known, on where a programme is most likely to fail. It asks five fixed questions and uses only public documents. It judges the plan on the page, not the people behind it.

The Bet

EU home affairs ministers endorsed a revised timeline on 5 March 2025 that placed ETIAS after the Entry/Exit System, with ETIAS expected in the last quarter of 2026 and transitional and grace periods of at least 12 months. The bet was that two new border systems could go live about a year apart. The design deserves credit: the fee is 20 euros, travellers under 18 or over 70 are exempt, and applicants are told to expect a decision within minutes. Those rules are published. What remains open is whether the organisation behind them can deliver to a date.

The Assumption

The timeline rests on one belief: that the Entry/Exit System would be stable before ETIAS needed to start. The evidence since is mixed. IATA says a standard passport check that once took 20 to 25 seconds now takes about 90, and a Financial Times and Qsensor study put average waits in Frankfurt at 120 minutes in July, against 60 a year earlier. The Commission’s spokesperson said on 1 September that the system had broadly worked well over the summer, and the temporary flexibilities that let countries suspend biometric collection at congested crossings were due to end on 6 September. Those are the conditions under which ETIAS would have had to launch.

The Sequence

The ETIAS quarter was endorsed in March 2025, seven months before the Entry/Exit System began operating. That system became fully operational on 10 April 2026, and the Commission’s explainer of 28 April still described ETIAS as scheduled for the last quarter of 2026, with the start date to follow later. On 9 July, The Local relayed a Financial Times report that the implementing agency had privately acknowledged a 2026 launch was no longer possible. Neither the Commission nor the agency had commented, and I found no confirmation since. The article also describes a letter to airlines in which Commissioner Magnus Brunner suggested most border problems stemmed from causes unrelated to the EES, such as staffing and infrastructure.

The Pager

When the requirement applies, the first person to meet a traveller without an authorisation will not be a minister. The Commission’s FAQ says carrier check-in staff and border guards can access each authorisation in the system, and that a visa-exempt traveller without a valid authorisation will be refused entry unless an exception applies. The same page says the EU will announce the start date several months before launch. It names no single office as accountable for that announcement. Check-in staff and border guards are the two roles that will apply the rule at the point of travel.

The Proof

Eighteen months after launch, the test is whether travellers can apply, receive a decision and board without incident at full volume. The Commission has said most applications will be decided within minutes, which is a measurable promise. The pages reviewed for this piece do not yet include a published target for how many applications will meet that standard, a ceiling for border queue times, a date by which either will be reported, or a named owner for that report. Success has no public definition yet, and the material reviewed shows no public commitment to a number.

Verdict

If the Commission publishes a start date, then it should publish alongside it the figures that would let the public judge the launch: the share of applications decided within a stated number of minutes, a maximum acceptable queue time at border posts, and a reporting date for both. If the date remains unpublished, then travellers and airlines should treat ETIAS as not in force, as the official FAQ itself describes it, and apply only through official channels once a date appears. A date announced several months ahead leaves time to define success first. A programme that defines success in advance can be held to it, and can claim credit when it meets it.

The CEO Called It Agility. The CFO Called It AI.

Posted on by

American Express confirmed on 23 October that it will cut roughly 5 per cent of its global workforce, alongside a $336 million pretax restructuring charge taken against third quarter results that beat expectations on both revenue and profit. Chairman and chief executive Stephen Squeri told American Banker the move was about speed and agility: “To accelerate innovation and win we need to become a more agile and efficient organization,” he said, adding that he had “asked the senior leadership team to streamline organization and reduce layers to speed decision making.” Nothing in that account named AI as a cause.

Hours later, on the same day, chief financial officer Christophe Le Caillec gave CNBC a different account of the identical restructuring: “So what we’re doing here is to reduce layers and increase spans of control, so that we can automate,” he said, tying the cuts explicitly to preparing the company to use AI “in the coming quarters and the coming years.” He added a caution worth sitting with on its own terms: “You don’t want to automate processes that are not optimally structured.”

 

Same Structural Change, Two Different Reasons Offered

Both executives described the identical mechanical change: fewer management layers, wider spans of control. What differs is which cause each man was willing to name in public, on the same day, about the same decision. Squeri’s version is one every board has heard before and rarely interrogates: agility, speed, efficiency. Le Caillec’s version names the one word Squeri’s own public remarks left out.

Neither man was lying. A restructuring can genuinely serve both purposes at once, and reducing layers to speed decisions is a real and defensible reason on its own terms. reporting on the cuts puts the reduction at around 5 per cent of a workforce numbered near 77,000, a scale that affects real people regardless of which executive’s framing a reader prefers. The interesting fact is not that the two accounts contradict each other outright. It is that they did not have to match, and nobody inside American Express seems to have thought they needed to.

 

The Test This Creates for Every Other Organisation

Most organisations making workforce cuts this year have picked one story and held it, usually the version without AI in it, since that framing invites fewer uncomfortable questions from employees, journalists and regulators alike. American Express’s two executives did not coordinate on a single line, and the gap between their answers is now public and permanent, sitting in two separate transcripts that anyone can compare.

That gap is a useful diagnostic for any PMO or governance function reviewing a workforce change tied to a technology programme. If a chief executive and a chief financial officer were each asked, separately, why a specific restructuring happened, would their answers converge on the same cause, or would the gap between them only become visible once someone bothered to check both transcripts. Most organisations have never actually tested this. They assume alignment because nobody has gone looking for the places it might not hold.

 

What a Governance Function Can Actually Check

A supportive PMO does not own the decision to restructure, and it should not try to. What it can own is making sure the organisation’s own public account of why a change happened is consistent before more than one senior leader is asked to give it. That is a cheap check, usually a single conversation before a results call or a town hall, and American Express’s own week shows what happens when nobody runs it: two true, defensible, individually reasonable answers that nonetheless read, side by side, like two different companies explaining two different decisions.

The deeper lesson sits past the embarrassment of a mismatched quote. If an organisation cannot agree internally on why it is cutting several thousand roles, the odds that it has agreed on how AI will actually be used inside the roles that remain are lower still.

Pick the last significant change your own organisation announced publicly. Ask two people who were in the room, separately, why it happened. Then compare the two answers you get.

AOMEI Knew Its Servers Were Compromised in June. It Still Hasn’t Told Everyone Affected.

Posted on by

AOMEI Backupper is backup software, the kind of tool procured once, configured once, and then forgotten, because forgetting about it is the point. Between April and July 2026, attackers altered two configuration files on AOMEI’s own update servers, down_html.ini, which tells the software where to download updates and what checksum to expect, and Notifycfg.ini, which controls the pop-up notifications that can trigger code execution. They supplied a matching checksum so the malicious package passed verification, then restored the original files within minutes. Confirmed tampering lasted roughly ten minutes on 28 May, six minutes on 16 July, and five minutes on 20 July. Any copy of the software checking for updates during one of those windows received a dropper called CROSSROAD, which in turn installed a legitimate remote access tool, Radmin, alongside a backdoor named SHINYROAD. On 20 July, a fourth technique tried to steal Windows password hashes by forcing machines to authenticate to an attacker controlled server, no malware required.

 

The Vendor Didn’t Believe Its Own Customers

ManageEngine’s endpoint detection software flagged the Radmin installation at two unrelated customers on 28 May. A customer raised the alert on 7 June, and ManageEngine told AOMEI that same day. AOMEI’s own account of what happened next is the part worth sitting with. It initially said the content had been intercepted and altered in transit, not on its own servers. ManageEngine disputed this, and said the evidence pointed to the vendor’s own delivery path. AOMEI investigated, said it had patched unspecified vulnerabilities, and could not identify a root cause. The attackers struck twice more, on 16 July and 20 July, while AOMEI believed the matter was closed. AOMEI only found traces of the intrusion after Mandiant and ManageEngine handed it the exact timestamps of the July windows. As of the most recent reporting, AOMEI had not identified or notified the customers whose machines were compromised.

This Is Not a Cybersecurity Story First

Treat this as an availability incident and the lesson is patch management. Treat it as what it actually is, a vendor trust failure in a category of software built entirely on trust, and the lesson changes. Backup software exists to be the thing an organisation reaches for when everything else has gone wrong. It runs with elevated privileges, it phones home for updates by design, and it rarely gets the scrutiny applied to software that actually touches customer data, precisely because it sits one layer removed from the systems everyone assumes carry the real risk. A PMO that tracks vendor risk at the point of procurement and never again has already missed where this incident happened. AOMEI was told directly, in June, that its update path had been compromised, and still could not confirm the problem was fixed before it recurred in July.

 

What a Governance Review Should Actually Ask

The question for any PMO or transformation office is not whether an organisation runs backup software from a reputable vendor. Every organisation does. The question is who owns the relationship with that vendor once the contract is signed, and whether that owner has a standing channel to hear about an incident like this one within days rather than months. Three things are worth checking this week. Does the vendor risk register include infrastructure tools that run with elevated privileges but rarely get named in a security review. Does the incident response plan assume a vendor will report promptly when its own update mechanism fails. Has anyone asked the backup vendor directly, this year, how it verifies the integrity of what it pushes to company machines.

 

The Software You Forgot About Is Watching Your Network Right Now

AOMEI has released a free detection tool and is urging customers to update and scan. That response arrived months after the first warning, and only after outside threat intelligence teams handed the vendor the exact minute its own servers were compromised. Pick the vendor in your own environment that nobody has reviewed since the contract was signed. Name who owns that relationship now, and ask whether they would know within a week if it failed the way AOMEI’s did.

The UK Government Cannot Say What Its Own Digital Transformation Costs

Posted on by

Dame Chi Onwurah chairs the House of Commons Science and Technology Committee. In October 2026 she asked the government a question that should have a straightforward answer: how much is Whitehall spending on digital transformation. The government’s written response admitted its financial systems cannot produce that figure at all.

“The government cannot provide a clear picture of how much is being spent on digital transformation,” Onwurah said when the committee published the response on 22 October. “It’s concerning that they cannot show where the money is going or how value for money is being measured.” Her verdict: “This must be addressed as a matter of urgency.”

The admission landed five months after the committee’s own report, Rewiring the state: Delivering digital government, set out what better visibility should look like. That June report recommended a target of 75% of digital spending funded through departmental resource budgets rather than capital budgets, up from an estimated 43 to 46% in 2021. It proposed that the Government Digital Service and the Treasury take explicit, named responsibility for hitting that target, reporting on progress every year against a performance framework GDS would have to publish first.

The government’s October response agreed to consider improvements, “over time.” It did not agree to appoint a Government Chief Digital Officer at permanent secretary level, the single accountable owner the committee wanted for exactly this kind of visibility gap. It also declined to name which departments and public bodies have not adopted basic data hygiene practices, the groundwork any spending framework would need to rest on.

 

A Governance Structure Cannot Report What It Cannot See

A supportive PMO exists to do one job before any delivery decision gets made: show the people with authority to act what is actually happening, in numbers they can trust. That reporting function is not decorative. It is the mechanism that lets a steering committee catch an overrun before it becomes unrecoverable, lets a sponsor compare one programme’s return against another’s, and lets a board ask a specific, answerable question about where money went.

Remove that function, and every layer of governance built on top of it is reporting on a picture it cannot actually see. Whitehall’s own select committee just demonstrated this at national scale. It is not that ministers are hiding the number. The government’s own position is that its financial systems were not built to produce it. A structure cannot be held accountable to a figure that structure itself cannot generate, and no amount of escalation, dashboarding or steering committee attendance changes that while the underlying number stays unavailable.

Visibility failures rarely announce themselves. Nobody in Whitehall woke up one day and decided spending should be untrackable. It accumulated, system by system, department by department, each decision locally sensible and collectively opaque, until a select committee had to ask the obvious question out loud before anyone could confirm the answer did not exist.

 

The Same Pattern Shows Up in Supplier Risk

The committee’s related finding on vendor concentration makes the same point from a different angle. It called reliance on a small number of suppliers, Palantir chief among them, an “unacceptable point of weakness” in public sector data handling, and recommended the government trigger the break clause in the £330 million NHS Federated Data Platform contract when it becomes available in 2027. The government has neither committed to nor ruled that out. A spending framework that cannot track its own cost is unlikely to have tracked supplier dependency any better, because both are the same discipline applied to different data: knowing what you have committed to, and to whom, before someone outside the organisation has to ask.

 

What a PMO Should Take From This

None of this requires a government department’s budget to test. Any PMO can ask itself the same question Onwurah put to Whitehall: if a board member asked for the total current spend across every active transformation initiative today, could that number be produced within a day, with a source for every pound in it. Few organisations answer yes immediately. Fewer still have a named owner accountable for making that answer better next quarter, the way the committee wanted GDS and the Treasury to be.

Gartner’s Own Survey Says 13% of CIOs Have Real AI Value. The Risk Is Winning Faster.

Posted on by

Gartner has spent two decades telling boards how to think about technology investment. On 19 October 2026, at its own IT Symposium/Xpo in Orlando, Gartner turned that measuring stick on itself and on AI, and the number it produced was 13%. That is the share of CIOs who say their enterprise has realised significant or transformational value from AI, according to Gartner’s 2027 CIO and Technology Executive Agenda survey of 2,002 CIOs and technology executives, fielded between May and July 2026.

Plenty of technology leaders already suspected AI’s return on investment was softer than the vendor decks claimed, so the 13% itself will surprise few of them. The real discomfort sits in what Gartner paired it with in the same week: a separate release titled “AI Risk is Rising Faster Than AI Value,” delivered as the opening keynote by Distinguished Analysts Daryl Plummer and Kristin Moyer. Gartner’s own research describes a gap between value and risk that is widening, on Gartner’s own watch.

 

The gap Gartner’s own numbers can’t explain away

Two figures explain the gap. Across Gartner’s research this year, 71% of organisations expect to deploy agentic AI within twelve months, yet only 21% have governance capabilities for it. In a separate Q3 2026 survey of 303 CIOs and IT risk leaders, only 25% said they actively align technology risk with business risk. Enterprises are building the engine faster than the brakes.

A widely cited number from Gartner’s research this year puts the risk-value gap even more starkly: 86% of CIOs believe AI risk is growing faster than AI value inside their own organisation. That statistic has circulated across several of Gartner’s 2026 Symposium events, from an earlier stop in Australia through to Orlando, and it measures something different from the 13% figure. The 13% is about value already realised. The 86% is about perceived trajectory. An organisation can report both honestly: little value banked so far, and risk accelerating ahead of whatever value eventually lands.

 

What Gartner’s own fix looks like

Plummer and Moyer’s answer took the shape of a governance structure they called an AI Central Bank, a function that holds value and risk accountable together, instead of letting a value team chase upside while a risk team chases the consequences separately. Moyer described it plainly: organisations can establish an AI Central Bank “to ensure transparency, coordination and accountability.” Plummer was equally direct about its limits: “What it will not do is eliminate risk, and it will not drive value,” he said. “The only thing that can do that, is people.” The central bank is scaffolding, making decisions visible and consistent across the enterprise. It does not make the decisions for anyone.

 

The vanguards did the unglamorous work first

That distinction matters more than the keynote language suggests, because the CIOs Gartner classifies as “vanguards” are the ones who built the unglamorous disciplines first, rather than the ones with the biggest AI budgets. Only 31% of surveyed CIOs have explicit criteria for killing an AI pilot before it reaches production, but those who do are 2.3 times more likely to be vanguards. Only 4% have plans to redesign roles for a mixed human-agent workforce, yet that group is 2.4 times more likely to report real value. Stopping criteria, cost accountability, workforce redesign: the pattern holds across every discipline Gartner measured. Governance is not what slows AI down. It is what lets the 13% become 20%.

 

What this looks like from the steering committee

None of this is abstract for anyone who has sat in a steering committee that approved an AI pilot without ever agreeing what success, or failure, would look like. The projects that die quietly in production are rarely killed by bad models. They are killed by the absence of a rule that says when to stop, and nobody in the room wanted to be the one who wrote it.

Meta and Tesla report third-quarter earnings the same week, and both face the same question from analysts that Gartner just put to CIOs directly: where is the AI value, and who is accountable for the risk beside it. Gartner’s own CIOs do not yet have a confident answer. Building the discipline to produce one is the job now.

Pre-Mortem: The VA Health Record Rollout and a $27 Billion Ceiling

Posted on by

The Department of Veterans Affairs approved a rise in the ceiling of its Oracle health-record contract from $10 billion to about $27 billion on 18 August, Stars and Stripes reported, and its next two medical centres, in Cleveland and Anchorage, go live on 24 October on VA’s own schedule. A ceiling sets the maximum VA may spend under the contract. The House Veterans’ Affairs Committee has subpoenaed Oracle’s executive chairman, Larry Ellison, and the chief executive of Oracle Health, Mike Sicilia.

A Pre-Mortem assumes a plan has failed and works backwards to find out why. The plan here is to replace a legacy record system in use since the 1980s with one commercial product, deployed in waves to roughly 170 medical centres by November 2031. Five fixed questions test it.

 

The Bet

VA’s bet is that one commercial record system, meant to be shared with the Department of Defense, can replace the legacy one, and that accelerating deployment after a pause is how it gets there. VA told Congress in September 2025 that finishing through 2031 would cost an estimated $37.2 billion, a spokesman said this month. The department also says the two-year pause added at least $1 billion in unexpected costs, and it cites unanticipated complexities for the larger ceiling. The prize is a single record for each veteran. VA signed the original deal with Cerner in 2018, and Oracle acquired Cerner in 2022.

 

The Assumption

The load-bearing belief is that the vendor would carry the cost of fixing its own system. The contract runs on a time-and-materials basis, so each piece of work is negotiated under the ceiling. In 2022 Sicilia told senators Oracle would fix performance and workflow problems at its own expense, Nextgov reported. At the 2 September hearing, Deputy Secretary Paul Lawrence said he had to establish what was promised and how VA’s acquisition team had treated it. The committee heard no detailed account from VA of what drove the extra $17 billion.

 

The Sequence

Credit is due for the restart. VA paused most deployments in April 2023, resumed in April 2026 with four Michigan hospitals, and publishes site names and dates through 2027. Eleven sites have gone live this year, and 26 more are scheduled for 2027. Then the order of events turns awkward. The ceiling was raised on 18 August, and committee members said the increase caught them off guard, although VA says it gave Congress the estimated total last year. The hearing that tested the increase came two weeks later. Oracle had accepted an invitation to testify and then declined.

 

The Pager

Pick one veteran with a referral booked at Cleveland the week after go-live. Who tells them the system is safe, and by what measure? VA says it tracks fifteen “return to normal” measures after each go-live, including emergency department throughput, surgeries and inpatient care, and that trouble tickets fall with each deployment. At the same hearing a House Democrat raised reports that about 300 dermatology referrals, some for patients with cancer diagnoses, were cancelled after a go-live. VA’s chief medical officer asked for the details so the department could investigate.

 

The Proof

The record will fill in quickly. Sicilia is due to testify on 19 November and Ellison on 10 December, according to the subpoenas. The Government Accountability Office says VA has not provided a cost estimate or detailed schedule documentation, so GAO cannot judge the schedule against good practice, and that 14 of its 18 recommendations, including all 12 priority ones, were not fully implemented as of August. Three tests are measurable: whether VA delivers a costed schedule, whether it publishes site-level results against its fifteen measures, and whether GAO’s priority recommendations close. Eighteen months, to April 2028, covers every 2027 wave.

 

Verdict

If VA delivers a costed schedule, publishes site-level results and closes GAO’s priority recommendations by April 2028, then the $27 billion ceiling bought a rollout that earned its pace, and other health systems can borrow the model. If the 26 sites of 2027 go live without them, then the ceiling bought speed and not proof, and the next cohort of veterans becomes the evidence. Any health system that puts one commercial record at the centre of national care owes its patients the same published test.

The $50 Million Question Every AI Vendor Contract Is Hiding

Posted on by

Read the indemnity clause in your AI vendor’s contract again. Then ask the vendor for the certificate of insurance sitting behind it. Most governance teams do the first and skip the second, and this month’s reporting on AI insurance explains why that second step is the one worth adding before your next renewal cycle.

 

 

The number behind the promise

The Wall Street Journal reported on 17 October that frontier AI developers are carrying potential legal liabilities running into the tens, and in some disclosed scenarios hundreds, of billions of dollars from safety incidents and wrongful-use claims. Specialist cover for those same companies tops out far lower. AIUC’s dedicated AI liability cover caps around $50 million per insured, Munich Re’s own aiSure product caps at roughly $15 million per claim, and limits from carriers such as Armilla and Testudo sit in a similar range. Insurers including AIG, W.R. Berkley, Great American, Chubb and Travelers have sought regulatory approval to strip AI-related claims out of standard liability policies altogether, and state regulators have approved the large majority of those requests. Some in the industry are now floating a federal liability backstop modelled on the Price-Anderson Act that underwrites nuclear power, though that is currently an industry request rather than a confirmed government commitment.

Separately, biotech entrepreneur Giorgos Tsetis argued in Fortune the same week that no individual carries personal accountability when an AI system causes harm. Companies absorb fines as a line item. The people who decided to ship the system, underfund the safety review, or ignore an internal warning carry none of that risk personally. Put his point next to the Journal’s numbers and the shape of the problem gets clearer: the financial backing behind an AI harm claim is thinner than advertised, and the personal accountability for causing that harm barely exists at all.

 

What an indemnity clause actually guarantees

An indemnity clause is a contractual promise to cover your losses if the vendor’s product causes harm. It is not evidence that the vendor can pay. A vendor can offer broad indemnity language and still carry a policy that caps at $50 million, a subsidiary structure that keeps liability away from the parent company, or no AI-specific cover at all because the general liability policy now excludes it by endorsement. None of that shows up in the clause. It shows up in the certificate of insurance, the policy schedule, and the corporate entity that actually signed the contract.

 

What this changes for PMO and governance review

For anyone standing up governance around healthcare transformation or health information exchange work, this is not an abstract insurance story, it is a contract due-diligence gap sitting inside your next vendor renewal. A contract touching patient data, clinical decision support, or an HIE integration carries exposure that runs into real money if the model gets something wrong: a misrouted referral, a hallucinated clinical summary, a leaked record set. The indemnity clause was written to make a risk committee comfortable signing off. Whether it can absorb a serious incident is a separate question, and until this month nobody had sized how wide that gap actually runs.

 

What to actually do about it

Three things change in how a PMO should run vendor due diligence from here.

Ask for the certificate of insurance and the stated policy limit before any AI-related contract reaches sign-off, not just the indemnity clause itself. If the vendor cannot produce one, treat the clause as marketing language until they can.

Check whether the policy names AI or algorithmic harm specifically, or is silent on it. A general liability policy written before the vendor’s AI product existed may not cover the product you are actually buying.

Size the gap against your own exposure. A $50 million cap set against a breach touching hundreds of thousands of patient records leaves the bulk of that exposure sitting with your organisation rather than the vendor’s.

This takes one new question added to the vendor risk template, asked at first sign-off and again at every renewal: show me the policy behind the promise, every time, before the signature goes on the page.

Pre-Mortem: Denmark’s Register Breach Was Caught by an Invoice

Posted on by

A small Danish company pays for every lookup it makes in the national population register, and in September someone used its account for more than 14 million lookup attempts. The bill, not a security alarm, exposed it on 2 October, according to The Next Web. The ministry says the lookups returned names, addresses and identity numbers for about 8.8 million records, on a register of roughly 11 million in a country of about six million, Bloomberg reported. Police have identified no one.

A Pre-Mortem assumes a plan has failed and works backwards to find out why. The plan here is the one most digital states run: let private firms with a legitimate interest query a national register, and trust each firm to guard its own door. Five fixed questions test it.

 

The Bet

Denmark bet that a national register can serve companies safely if each company carries its own security duties. That design is efficient for the state: one register, many users, each answerable for its own account. Under the CPR Act, firms receive data on people they have already identified one by one, and the ministry sets the access terms, including security measures. The register’s administration describes the incident as the misuse of a company’s lawful access, and says the data excluded the names and addresses of people with name-and-address protection. Monitoring sat with every firm holding an account. The prize sat with the register.

 

The Assumption

The load-bearing belief is that a firm will notice when its own account is abused, and that the register will notice if it does not. Documents obtained by TV 2 and reported by NordiskPost show that unauthorised access began on 10 September and appears to have stopped around 20 September. It was detected on 2 October, twelve days later. The account stayed open for more than 21 days. Minister Christina Egelund has acknowledged that the security arrangements were insufficient and ordered a review with no deadline. The documents leave open why 14 million lookups, far above the firm’s normal activity, triggered no alert.

 

The Sequence

Credit is due for the pace after discovery. Officials flagged the invoice on the evening of 2 October, The Next Web reports, and officers from the National Unit for Special Crime visited the company the next evening. The ministry announced the breach on Monday 5 October and briefed the press the following day. Then the order of events turns awkward. On 9 October the register’s administration wrote to 1,039 companies reminding them to appoint a security lead, review staff access and monitor the number of searches on their accounts. Those duties already existed. The reminder arrived after the breach, and the breach itself surfaced through a bill.

 

The Pager

Pick the pharmacist at the counter who accepts a ten-digit number as proof of who is standing there. The register’s own guidance already says the number must not be the only proof of identity, yet pharmacies accept it alone. The head of Denmark’s agency for civil security has now told firms and authorities to stop, and pharmacies must find another way. Ordinary citizens ask something simpler: if my number is out, can I get a new one? Egelund has said it is too early to know, although the law allows a replacement in special cases where a number has been misused. A guideline that went unenforced became an instruction only after the number was public.

 

The Proof

The facts that would settle this are still open. Datatilsynet, the data protection authority, had published no assessment of the case by 13 October. The security review has no end date. Police have not publicly identified the attacker or what became of the data. Three tests are measurable: whether Denmark publishes the review, whether it puts volume alarms and limits on every account that can query the register, and whether leaked numbers can be replaced. Eighteen months, to April 2028, is enough time to see whether any of the three happened or whether the register still depends on the invoice.

 

Verdict

If the review ends with automatic volume alarms, per-account limits and a way to replace a leaked number, then delegated access survives and Denmark’s register becomes a template for other states. If it ends with reminder letters, then the next abuse will be found by an invoice as well, and a national register will keep depending on the weakest of more than a thousand companies. Any country that lets firms query a population register shares that exposure.

Resistance Is the Data

Posted on by

When a technology programme meets resistance, the standard response is to escalate the change effort: more communications, more training sessions, more stakeholder workshops. The assumption underneath all of it is that resistance is the obstacle, something to be persuaded, equipped or managed past.

That assumption is usually wrong, and acting on it tends to make the programme worse. Resistance to a technology implementation is not noise in the signal. It is the signal.

 

What Resistance Is Actually Telling You

BCG’s 2020 survey of 825 senior executives found that 70 per cent of digital transformations fall short of their objectives. Bain’s 2024 Transformation and Change Survey of more than 400 executives put the figure higher still, at 88 per cent failing to achieve their original ambition. The two surveys measure different things, but both point the same way. Those programmes were rarely undone by stubborn employees. Their implementation drifted from operational reality, and the people doing the work could see it before the programme team could.

When a clinical team pushes back on a new patient records system, they are rarely resisting technology itself. They are reporting that the workflow design does not match how the ward runs at three in the morning, or that the training schedule assumes competencies the ward does not yet have. A finance team resisting a new reporting platform is often reporting that it strips out a nuance the team relies on daily, one the requirements process never surfaced.

The people closest to delivery are often the sharpest diagnosticians of implementation risk a programme has, and under schedule pressure, programme teams tend to receive that diagnosis as obstruction rather than intelligence.

 

The Stakeholder Management Euphemism

The phrase stakeholder management carries a weight of assumptions worth questioning. In practice, it frequently means: how do we get these people to comply?

The vocabulary gives it away: stakeholders segmented by resistance level, engagement plans that move people along an adoption curve, communications calibrated to reduce objections. The apparatus treats resistance as something to overcome rather than something to process.

Cornerstone OnDemand’s 2024 workforce readiness research found that 63 per cent of UK executives regard their own workforce as unprepared for change, and that employees without proper access to training were five times more likely to report difficulty adapting. That gap is a programme design failure. The workforce did not create it.

Whatfix’s survey of more than 1,100 software users found that 51 per cent had no role-based onboarding or support when a new system went live, and 78 per cent felt they lacked the expertise to use their daily tools properly. The investment did not happen. The resistance did.

 

What Good CIOs Do Differently

The best technology leaders share a habit: they treat early resistance as a discovery mechanism rather than a threat. When a pilot site pushes back, the first useful question is what the pilot site is seeing that the programme team is not, rather than how to handle the resistance.

That reframe slows the rollout and demands real engagement, but it catches design flaws at pilot cost instead of full-deployment cost. Decades of change research have produced one widely repeated adoption-failure statistic after another, none traceable to a verifiable source. The underlying point holds regardless: resistance is useful information, whatever number gets quoted beside it.

 

Change Fatigue and the Portfolio Problem

Reading resistance accurately gets harder inside an organisation running too many programmes at once. Prosci has benchmarked change saturation since 2007, and its latest research found 73 per cent of organisations near, at or past that point.

That is a portfolio problem wearing a change management label. Resistance to one programme, under those conditions, is often resistance to being asked to run faster on a track that is already full.

 

Reading the Room Before It Becomes a War Room

Before a programme leaves design, ask the people most affected a blunt question: what about this will not work? The phrasing matters. “What concerns do you have?” invites the polite version of resistance. “What will not work?” leaves no room for one.

Asked honestly, that question surfaces operational objections while they are still cheap to fix. Most of the time, the people who will resist a programme know it has problems before the programme team does. The only real question is whether the programme is structured to hear them, or finds out after go-live, when the resistance has nowhere left to go but into next year’s failure numbers.