Artificial intelligence has quickly moved from being an emerging technology to becoming part of everyday business operations. From writing content and analysing data to assisting research, automating routine processes and supporting customer interactions, AI tools are changing how individuals and organisations approach their work. Tasks that once required hours of manual effort can now be completed in considerably less time.
For businesses, this has created opportunities to improve productivity, manage operations more efficiently, and explore new ways of serving customers. For individuals, AI has become an accessible tool for finding information, learning new skills, and simplifying everyday activities.
However, as adoption continues to grow, the global conversation around artificial intelligence is beginning to change. Alongside its capabilities, questions about its reliability, financial viability, safety, and accountability are receiving greater attention.
Recent developments involving some of the world’s largest technology companies, legal challenges against AI developers, concerns from the scientific community, and changing investor sentiment suggest that the technology is entering a more closely scrutinised phase. The discussion is no longer limited to what AI can accomplish. It increasingly involves how responsibly it is developed, how accurately it performs, and whether its commercial promise can translate into sustainable value.
Questions About Returns about AI Investment is Growing
One of the clearest indications of this changing perception is evident in the financial markets.
Technology companies have committed substantial resources to developing AI infrastructure, expanding data centre capacity and building increasingly advanced models. However, investors are now paying closer attention to whether these investments are generating returns at the pace the market expects.
Alphabet’s (Google’s parent company) latest quarterly results brought this concern into focus. The company reported negative free cash flow of $9.07 billion for the third quarter of 2026, significantly missing analysts’ expectations of $2.7 billion. Despite maintaining its full-year capital expenditure guidance of $195 billion to $205 billion, the company’s shares fell nearly 13% in a single trading session on October 29, wiping out approximately $500 billion in market value.
The company’s search advertising business continued to grow, recording a 12% increase in revenue to $63.35 billion. However, it fell short of analysts’ expectations. Alphabet also acknowledged that monetisation within its AI Mode remained lower, even as it expanded advertising across AI Overviews and other AI-powered search experiences. (Source: CNBC)
The development highlights a challenge facing established technology companies. While they are investing heavily in AI to remain competitive, they must also protect existing revenue streams and demonstrate that their new offerings can support long-term business growth.
A similar trend has emerged in China, where investor enthusiasm around AI and semiconductor companies has faced a sharp correction. According to a Reuters report published by The Economic Times, China’s STAR50 index, which tracks major AI and semiconductor companies, declined by 40% in less than four months, wiping out approximately $480 billion in market value. (Source: Economic Times)
The decline has been linked to concerns around high valuations, demand for AI technology and the sustainability of investment in the sector. Omar Agamy also explains this development in his Instagram reel, highlighting what China’s AI market correction reveals about the gap between investor expectations and the actual financial performance of AI companies.
These developments do not necessarily indicate that AI’s long-term potential is diminishing. Instead, they reflect a growing distinction between technological progress and the commercial returns expected from it. For business leaders, the question is increasingly about how AI investments translate into measurable productivity, revenue and operational value.
When AI’s Convenience Meets the Question of Reliability
While investors are examining the financial side of AI, concerns about its everyday applications are also gaining attention. AI tools are increasingly used to retrieve information, generate recommendations, support research and assist with decisions. However, their accessibility and conversational nature do not always guarantee accurate or appropriate responses.
A study published in August 2025 by the Center for Countering Digital Hate (CCDH), reported by the Associated Press and PBS NewsHour, examined how ChatGPT responded to prompts involving teenagers and sensitive subjects. Researchers found that the chatbot could provide problematic, personalised responses even when conversations initially included warnings against risky behaviour. More than half of the 1,200 responses examined were classified as dangerous. (Source: Congress GOV)
The study also raised questions about the effectiveness of AI safety measures, as researchers were able to obtain responses to restricted requests by changing the context of their prompts. Unlike search engines that present links to different sources, chatbots combine information into personalised responses. While this makes them useful, it can also make inaccurate guidance appear convincing.
These concerns have also entered the legal arena. In a wrongful-death lawsuit filed in California, the parents of 19-year-old Sam Nelson alleged that ChatGPT provided guidance encouraging their son to combine dangerous substances before his death from an overdose in May 2025. The complaint further claimed that the chatbot’s responses became more permissive following the launch of GPT-4o in 2024. (Source: Times Now News)
OpenAI described the death as heartbreaking and said it had continued strengthening its safeguards for sensitive situations. The company also stated that the model involved was no longer available. While the allegations have not been established as findings of legal liability, the case raises questions about the responsibilities of AI developers when users seek guidance on health and safety matters.
A separate case reported by The New York Times highlights the risks of relying on AI for personal medical decisions. Bowen Wan, a 31-year-old woman with epilepsy, increasingly turned to ChatGPT for health-related advice. According to the report, she followed a medication-reduction plan suggested by the chatbot without consulting her neurologist. Her parents later sued OpenAI, alleging that the chatbot encouraged her to reduce her medication and reassured her about symptoms that required medical attention. Her death was officially attributed to sudden unexpected death in epilepsy (SUDEP); the lawsuit raises questions about whether the chatbot’s guidance played a role in the events leading up to it.
A reel by NYT Well also discusses the case, highlighting the risks of treating AI-generated health guidance as a substitute for professional medical advice.
For users, these cases underline why AI-generated information should not automatically be treated as professional advice. For businesses developing and deploying these tools, they highlight the need to test model behaviour across complex interactions and ensure that appropriate safeguards are in place.
AI Is Also Challenging Established Professional Practices
The debate around AI reliability extends beyond consumer applications and sensitive conversations. Its growing presence in specialised fields is also prompting discussions about professional standards, research ownership and accountability.
In October 2026, the Association for Human Mathematics called for a boycott of OpenAI after the company released findings involving 377 mathematical problems, alongside a repository containing more than 700 files. (Source: India Today)
The association questioned the company’s approach to AI-generated mathematical research, arguing that the work had not been requested by the mathematical community. It also raised concerns about whether AI-generated mathematical arguments could be independently understood, verified and accepted within established research practices.
An independent Advisory Group on Mathematics and Artificial Intelligence has similarly highlighted the difficulty of evaluating complex mathematical arguments produced by advanced AI systems.
OpenAI, meanwhile, shared formal versions of several proofs using Lean, a programming language that allows mathematical arguments to be checked computationally. The company also released reasoning summaries and details about its computational approach.
The disagreement illustrates a challenge that will become increasingly relevant as AI enters specialised professional environments. Producing an answer is one part of the process; establishing that the answer is correct, transparent, and independently verifiable is another.
There are also concerns about unequal access to advanced AI systems. Researchers working within technology companies may have access to computing resources and models that independent professionals cannot easily obtain.
For businesses adopting AI in research, finance, legal services and other specialised functions, this distinction matters. Efficiency gains are valuable, but the ability to verify outputs and retain human accountability remains essential.
What This Changing Perception Means for Businesses
Taken together, these developments point towards a more balanced understanding of artificial intelligence. Its capabilities are expanding, but so is the need to assess where it performs well, where it requires human intervention, and where its limitations could create risks.
For businesses, this calls for a more considered approach to AI adoption.
- First, productivity should be measured against actual business outcomes. Introducing AI into an operation does not automatically make it more efficient. Organisations need to evaluate whether the technology reduces costs, improves turnaround times, supports employees, or contributes to better customer experiences.
- Second, human oversight should remain part of important decisions. AI can help teams analyse information, prepare recommendations and process large volumes of data. However, outputs involving financial commitments, legal matters, healthcare or customer-sensitive decisions require appropriate verification and human judgement.
- Third, businesses need to understand the limitations of the tools they deploy. Different AI systems have different capabilities, data dependencies, and accuracy levels. Organisations should evaluate their performance in real operating conditions rather than relying solely on product demonstrations or promised capabilities.
- Finally, responsible adoption must extend beyond the initial implementation. Regular testing, data protection, clear accountability, and employee awareness should form part of an organisation’s AI strategy. As models evolve through updates, their performance and behaviour also need to be reviewed.
The Next Phase of AI Will Depend on More Than Innovation
Artificial intelligence is unlikely to disappear from business or everyday life. Its ability to process information, support complex tasks and improve productivity will continue to create opportunities across industries. However, the developments unfolding across global markets, technology companies and professional communities suggest that innovation alone will not determine how successfully AI becomes part of the wider economy.
Financial sustainability, reliability, transparency and public trust will also influence its adoption. Companies that develop AI will need to demonstrate not only what their systems can do but also how they manage the risks associated with them. Businesses using these tools will need to understand both their capabilities and their limitations. For individuals, the ability to use AI effectively will increasingly involve knowing when to rely on it and when to question its responses.
The more useful conversation, therefore, is not whether AI is good or bad for businesses and society. It is about understanding the conditions under which it delivers value and the safeguards needed when it does not.
AI can make work faster, but speed alone is not a measure of progress. Its long-term value will depend on how accurately, responsibly and effectively people and organisations learn to use it.













