Deloitte has launched one of the most expansive enterprise AI rollouts to date, announcing a major expansion of its partnership with Anthropic to deploy the Claude AI platform across its entire global workforce of 470,000 employees. The firm plans to embed Claude into workflows spanning finance, healthcare, and public services, while also training 15,000 specialists and establishing a Claude Center of Excellence to guide adoption and governance.
Anthropic describes the deal as its largest enterprise deployment so far, cementing Deloitte’s strategic commitment to AI at scale.
However, the announcement’s fanfare was quickly complicated by a parallel story out of Australia.
On the same day the Claude partnership was revealed, Deloitte confirmed it would partially refund the Australian government after delivering a flawed report riddled with AI-generated errors.
The 237-page report commissioned by the Department of Employment and Workplace Relations for AU$439,000 contained fake citations, misquoted legal rulings, and references to academic work that doesn’t exist.
A revised version has since been published along with a disclosure that Azure OpenAI was used in drafting the report.
Why It Matters: This dual development captures the promise and peril of enterprise AI adoption. Deloitte is positioning itself as a global AI leader, but its misstep in Australia demonstrates just how fragile that position can be without stringent quality controls. As generative tools move deeper into regulated domains, errors are now high-stakes public liabilities.
- Claude to Power AI Across Deloitte’s Global Workforce: Deloitte’s expanded alliance with Anthropic will bring the Claude assistant to all 470,000 of its global employees. This includes plans to develop compliance tools for highly regulated industries, as well as the creation of department-specific “personas” for internal use, such as agents tailored for accountants or software developers. Deloitte will also open a Claude Center of Excellence to standardize implementations and share best practices across its network.
- Formal Training and Certification for 15,000 Staff: In a bid to professionalize AI usage, Deloitte and Anthropic are co-developing a certification program to train 15,000 employees on Claude. These certified professionals will lead implementations for both internal operations and external client engagements, signaling a deliberate strategy to scale responsibly while maintaining technical and ethical standards.
- Government Refund Over AI-Hallucinated Report: On the same day Deloitte announced its Claude partnership, Deloitte Australia agreed to refund part of an AU$439,000 contract with the Department of Employment and Workplace Relations. The report looked at how automated penalties were used in the country’s welfare system and was found to include made-up quotes, fake academic references, and incorrect attributions, with one quote wrongly linked to a federal court judgment. A corrected version was released last week with the false information removed and a note disclosing that a language model had been used to help write the report.
- AI’s Role Acknowledged, But Details Remain Opaque: The revised report confirms that Azure OpenAI was used to help write the original version, but Deloitte has not explicitly stated whether the hallucinations were produced by the AI model. The company told the Associated Press that the matter had been “resolved directly with the client” but declined to comment further. Senator Barbara Pocock of the Australian Greens criticized Deloitte’s approach as “the kind of thing a first-year university student would be in deep trouble for,” and argued the firm should refund the entire amount.
- A Broader Pattern of AI-Generated Inaccuracies: Deloitte’s incident reflects a broader pattern of mistakes involving language models across different industries. Earlier this year, Anthropic faced criticism after Claude produced a false legal citation during a copyright dispute. Amazon’s internal tool reportedly struggled with accuracy, and a Chicago newspaper published a book list that included made-up titles.
Deloitte to partially refund Australia for report with apparent AI-generated errors – ABC News
Deloitte will make Claude available to 470,000 people across its global network – Anthropic
Trusted insights for technology leaders
Our readers are CIOs, CTOs, and senior IT executives who rely on The National CIO Review for smart, curated takes on the trends shaping the enterprise, from GenAI to cybersecurity and beyond.
Subscribe to our 4x a week newsletter to keep up with the insights that matter.
