Analysts warn most companies build AI on weak data and unclear oversight

Industry researchers reported that most companies buy AI tools before fixing the business data those tools depend on. A 2025 MIT study of 300 company AI projects found that about 95 percent produced no measurable financial gain. Adobe governance director Joseph Wallace wrote that most large firms set AI rules but give no one clear power to stop a harmful AI system.

Security researcher Ian Carroll said he used Anthropic's Claude AI in April to exploit a flaw in Front Gate Tickets. The platform, owned by Live Nation, handles ticketing for major US festivals such as Lollapalooza and Bonnaroo. Carroll said the flaw let him issue unlimited free tickets, including a $4,000 pass, but he reported it and it was fixed within a day.

Franklin Templeton, an investment firm that manages about $1.66 trillion, said it is moving AI from small tests into everyday work across the company. The firm uses AI systems in research, operations, sales, and compliance, working with a technology partner called Wand AI. Franklin Templeton said human staff keep oversight, with named owners for each AI system.

Bank of England deputy governor Sarah Breeden said on June 30 that current rules were not built for AI that trades on its own. At a European Central Bank forum in Portugal, she said the bank is studying new safeguards, including a market-wide "kill switch." The tool could stop trading if many AI systems fail at once, and she said more than half of finance firms already use such systems.

Getty Images said its board voted on June 30 to end its planned $3.7 billion merger with rival Shutterstock. The company walked away after the UK competition regulator required Shutterstock to sell its editorial photo business as a condition for approval. About nine days earlier, Getty signed a deal with OpenAI to show its licensed images inside ChatGPT.