THE
ACCOUNT
The latest in finance and business
THE
ACCOUNT
The latest in finance and business
Danger zones
Each year the World Economic Forum produces its Global Risks Report. Admittedly, it doesn’t make for cheery reading, but it’s fascinating nonetheless.
The top 10 short and long-term risks are shown in the accompanying graphic. No doubt, you will have your own interpretations of what they say about the world today.
Misinformation and disinformation, which surged to number one on the list of short-term threats in 2024 and 2025, fell to number two. Not, we assume, because big tech firms have tightened controls on fake news (far from it), but because geometric confrontation, which was ranked ninth in 2025, surged to number one on the back of Trump, tariffs and trade disputes sporadically dominating the headlines through the year.
The responses were collected from 12 August–22 September 2025 from 1,300 leaders in business, government and other organisations.
While the “adverse outcomes” of AI technologies are placed fifth on the list of long-term risks, you might wonder whether the possibility of its investment bubble bursting may have featured had those responses been submitted a few months later.
Danger zones
Each year the World Economic Forum produces its Global Risks Report. Admittedly, it doesn’t make for cheery reading, but it’s fascinating nonetheless.
The top 10 short and long-term risks are shown in the accompanying graphic. No doubt, you will have your own interpretations of what they say about the world today.
Misinformation and disinformation, which surged to number one on the list of short-term threats in 2024 and 2025, fell to number two. Not, we assume, because big tech firms have tightened controls on fake news (far from it), but because geometric confrontation, which was ranked ninth in 2025, surged to number one on the back of Trump, tariffs and trade disputes sporadically dominating the headlines through the year.
The responses were collected from 12 August–22 September 2025 from 1,300 leaders in business, government and other organisations.
While the “adverse outcomes” of AI technologies are placed fifth on the list of long-term risks, you might wonder whether the possibility of its investment bubble bursting may have featured had those responses been submitted a few months later.
Training day
Half of the UK’s mid-sized companies plan to increase the number of apprentices they recruit, according to a recent survey carried out by BDO. The findings, compiled from responses from 500 business leaders, show that 26% plan to begin taking on apprentices this year, while 24% aim to increase their existing intake. The increases follow the government’s announcement of free training for apprentices under 25.
Furthermore, that trend goes across the board, albeit with some sectors leading the way. Construction (56%), technology (55%) and financial services (54%) are planning to either start or expand their apprenticeships numbers this year.
If they follow through on those projections, it will come as a huge relief to school leavers, who are increasingly at risk of seeing entry-level roles disappear, as the first rung of the career ladder is supplanted to AI tools.
“It’s encouraging to see so many mid-sized businesses, the engine of the UK economy, doubling down on their apprenticeship hiring, despite challenging economic conditions,” said BDO partner Richard Austin. “A national programme linking local businesses with local schools and colleges would be a great next step."
Training day
Half of the UK’s mid-sized companies plan to increase the number of apprentices they recruit, according to a recent survey carried out by BDO. The findings, compiled from responses from 500 business leaders, show that 26% plan to begin taking on apprentices this year, while 24% aim to increase their existing intake. The increases follow the government’s announcement of free training for apprentices under 25.
Furthermore, that trend goes across the board, albeit with some sectors leading the way. Construction (56%), technology (55%) and financial services (54%) are planning to either start or expand their apprenticeships numbers this year.
If they follow through on those projections, it will come as a huge relief to school leavers, who are increasingly at risk of seeing entry-level roles disappear, as the first rung of the career ladder is supplanted to AI tools.
“It’s encouraging to see so many mid-sized businesses, the engine of the UK economy, doubling down on their apprenticeship hiring, despite challenging economic conditions,” said BDO partner Richard Austin. “A national programme linking local businesses with local schools and colleges would be a great next step."
Click caution
Cybercrime researchers have identified a new ruse being used to trick potential victims into handing over remote control of their devices – and they’re calling it the ‘Voicemail Trap’.
It involves sending a notification to a phone, instructing the user to click a link to hear an urgent recorded voice message.
According to Black Cyber Security, “The messages often impersonate trusted financial organisations and direct recipients to convincing, bank-themed websites. Victims are told to download an ‘audio update’ to hear the message, but the file is a script that silently installs legitimate remote management software, allowing attackers persistent access to steal data or deploy further malware.”
The scam even involves playing an audio file to trick the user into believing the the update was successfully completed.
“Researchers observed 86 websites linked to this activity on 12 January 2026,” added Black Cyber Security. “Leaders should reinforce click caution and block untrusted download prompts.”
Click caution
Cybercrime researchers have identified a new ruse being used to trick potential victims into handing over remote control of their devices – and they’re calling it the ‘Voicemail Trap’.
It involves sending a notification to a phone, instructing the user to click a link to hear an urgent recorded voice message.
According to Black Cyber Security, “The messages often impersonate trusted financial organisations and direct recipients to convincing, bank-themed websites. Victims are told to download an ‘audio update’ to hear the message, but the file is a script that silently installs legitimate remote management software, allowing attackers persistent access to steal data or deploy further malware.”
The scam even involves playing an audio file to trick the user into believing the the update was successfully completed.
“Researchers observed 86 websites linked to this activity on 12 January 2026,” added Black Cyber Security. “Leaders should reinforce click caution and block untrusted download prompts.”
Money for nothing
Achieving ‘unicorn’ status – when a privately held new business hits a $1bn-plus (£730m) valuation – used to be the ultimate ambition of any start-up. But that is so 2010s. Last year, a company founded in February raised a world record $2bn, at a total valuation of $12bn. But here’s the punchline. This business didn’t even have a name, let alone a product to justify all that seed capital.
The person behind the start-up, which was eventually named Thinking Machines, is Mira Murati, formerly CTO at OpenAI. She assembled a leadership team packed with her former employer’s senior colleagues. That alone was sufficient to generate enough buzz among VCs and other tech companies that the usual suspects when it comes to mega-investments in tech start-ups (Nvidia, Andreessen Horowitz, Cisco, Accel and more) rushed to get involved. There was even £10m from the government of Albania, seemingly caught up in the hype cycle, mixed with a spot of national pride that Murati herself is Albanian by birth.
In October, Thinking Machines launched Tinker, an AI that takes the vast amount of data fed into large language models (LLMs), such as ChatGPT, Claude and Gemini, and fine-tunes it for more specific use. In theory, it could drive efficiencies for companies that input their own data, along with all the other data available, to deliver a bespoke solution to their business.
But here’s the catch. Tinker uses information from open-source LLMs such as DeepSeek, which anyone can access. ChatGPT, Gemini and Claude are closed source, and fuelled by vast sums of investment that eventually need to deliver a profit. Therefore, they cannot be accessed by the public. What’s more, ChatGPT and Gemini make up over 80% of all AI traffic outside of China, meaning Tinker will be trained on much smaller data sets that are likely to be of less relevance to US users.
Then, in November, Thinking Machines started seeking more investment, in line with Murati’s claim that $50bn was a more realistic value for her company. At the time of writing, the investment has not been forthcoming. Meanwhile, more than half the senior team have been poached by other companies, such as Meta, while co-founder, Barret Zoph, was fired, reputedly for an undisclosed affair with another member of staff. Zoph, who denied the allegation, opted to return to his previous employer, OpenAI.
Thinking Machines may yet prove to be a success. But right now Tinker looks like a turkey – and the most extreme example yet of how AI can consume vast quantities of venture capital based on little more than hype, while giving very little in return.
Ryan Herman
Money for nothing
Achieving ‘unicorn’ status – when a privately held new business hits a $1bn-plus (£730m) valuation – used to be the ultimate ambition of any start-up. But that is so 2010s. Last year, a company founded in February raised a world record $2bn, at a total valuation of $12bn. But here’s the punchline. This business didn’t even have a name, let alone a product to justify all that seed capital.
The person behind the start-up, which was eventually named Thinking Machines, is Mira Murati, formerly CTO at OpenAI. She assembled a leadership team packed with her former employer’s senior colleagues. That alone was sufficient to generate enough buzz among VCs and other tech companies that the usual suspects when it comes to mega-investments in tech start-ups (Nvidia, Andreessen Horowitz, Cisco, Accel and more) rushed to get involved. There was even £10m from the government of Albania, seemingly caught up in the hype cycle, mixed with a spot of national pride that Murati herself is Albanian by birth.
In October, Thinking Machines launched Tinker, an AI that takes the vast amount of data fed into large language models (LLMs), such as ChatGPT, Claude and Gemini, and fine-tunes it for more specific use. In theory, it could drive efficiencies for companies that input their own data, along with all the other data available, to deliver a bespoke solution to their business.
But here’s the catch. Tinker uses information from open-source LLMs such as DeepSeek, which anyone can access. ChatGPT, Gemini and Claude are closed source, and fuelled by vast sums of investment that eventually need to deliver a profit. Therefore, they cannot be accessed by the public. What’s more, ChatGPT and Gemini make up over 80% of all AI traffic outside of China, meaning Tinker will be trained on much smaller data sets that are likely to be of less relevance to US users.
Then, in November, Thinking Machines started seeking more investment, in line with Murati’s claim that $50bn was a more realistic value for her company. At the time of writing, the investment has not been forthcoming. Meanwhile, more than half the senior team have been poached by other companies, such as Meta, while co-founder, Barret Zoph, was fired, reputedly for an undisclosed affair with another member of staff. Zoph, who denied the allegation, opted to return to his previous employer, OpenAI.
Thinking Machines may yet prove to be a success. But right now Tinker looks like a turkey – and the most extreme example yet of how AI can consume vast quantities of venture capital based on little more than hype, while giving very little in return.
Ryan Herman
Built with Shorthand