AI in the deal room
What Deutsche Bank Reveals About AI in the Deal Room
Siân Arkan · 27 October 2026
The Wrong Case
AI can analyse thousands of documents in seconds. But what happens when the information it produces is completely wrong?
Imagine preparing for a major transaction or case. The documents are extensive, the deadline is approaching, and artificial intelligence (AI) promises to make the process easier. It identifies relevant legal authorities and presents its findings with confidence. Everything appears ready in a fraction of the time it would have otherwise taken (without AI). There is just one problem: some of the cases it cites do not exist.
This is not a hypothetical scenario. In September 2026, the District of Columbia Court of Appeals issued a published order in Douglas v Deutsche Bank National Trust Co., highlighting the dangers of relying on AI without proper verification. Although the incident arose in litigation rather than a commercial transaction, it raises an important question for the legal deal room: are efficiency and convenience coming at the expense of accuracy and professional judgement?
When AI gets the law wrong
The case concerned a mortgage foreclosure appeal involving Deutsche Bank National Trust Company. While reviewing the bank's legal brief, the court discovered references to several cases it could not verify. The lawyer responsible subsequently confirmed that four cited authorities did not exist.
She admitted that she had used Google's generative AI search tool: ‘Gemini’ to assist with locating legal authorities. However, she had failed to check the existence or accuracy of the citations before filing the brief. The court struck the bank's brief and referred the matter to the Office of Disciplinary Counsel for any investigation it considered appropriate. The court also criticised the failure to verify the authorities. Importantly, the referral did not itself establish that the lawyer had committed professional misconduct.
The problem was not simply that AI had generated inaccurate information. It was that the information had been presented as legitimate legal authority without being checked. What appeared to be a shortcut had become a serious professional error.
The deal room's hidden vulnerability
Deal rooms are built around information. During mergers and acquisitions (M&A), lawyers and financial advisers may examine thousands of important documents to identify liabilities and determine whether a transaction should proceed.
AI offers an attractive solution to this demanding process. It can help review large volumes of material, summarise lengthy contracts and flag clauses that require closer attention. Used appropriately, it can free professionals to concentrate on strategic negotiation and complex legal analysis.
However, the Deutsche Bank incident exposes a less obvious danger: inaccurate information can appear sufficiently convincing to escape scrutiny.Imagine an AI tool reviewing a company's contracts before an acquisition. It incorrectly summarises a termination clause or overlooks a restriction on transferring a contract to a new owner. If the deal team accepts that summary without consulting the original document, its assessment of the target company could be fundamentally flawed. The buyer might complete the transaction only to discover unexpected liabilities or commercial restrictions that undermine the value of the acquisition.
Unlike an obvious error, an AI-generated mistake can look entirely credible. This makes verification particularly important when decisions depend on the smallest contractual details.
Who takes responsibility?
The central issue is accountability. If an AI tool produces incorrect information, should responsibility fall on the software provider, the law firm or the professional who relies on its output?
AI may contribute to an error, but it does not remove a lawyer's professional obligations. The Deutsche Bank case illustrates that using AI does not excuse a failure to verify legal authorities. The same principle applies to transactional advice: professionals must understand their tools' limitations and exercise independent judgement.
There is also a wider commercial consideration. Clients pay for expertise, reliability and informed advice, not simply the fastest possible review of their documents. If AI-assisted work is accepted without adequate checks, the damage may extend beyond financial losses to reputational harm and a breakdown of client trust. The answer is not necessarily to prohibit AI. Instead, firms need clear policies, appropriate training and review processes that ensure AI-generated findings are checked against reliable sources and original documents.
Efficiency is not the same as expertise
AI has considerable potential to improve legal work, and rejecting it entirely would mean overlooking genuine opportunities. But the Deutsche Bank incident demonstrates why technological capability must not be confused with professional competence.
The future of the deal room will depend on finding the right balance between automation and human oversight. AI can help identify information, but professionals must assess its accuracy, understand its significance and take responsibility for the advice they provide.
Ultimately, the most valuable skill in an AI-assisted deal room may not be knowing how to generate an answer. It may be knowing when that answer cannot be trusted.
Because when millions of pounds and commercial relationships are at stake, a convincing answer is not enough. It must also be correct.
This is a general explainer on industry trends, not advice on any specific tool, firm or transaction, and it isn't legal advice.