Simply sign up to the Investments myFT Digest — delivered directly to your inbox.
Advertising agency WPP has struggled to stem an exodus of customers at the same time that it is tailoring its business model to address the changes which have been wrought across the industry through digitalisation and artificial intelligence.
Takeover speculation gave the company’s shares a small boost this week, but its market valuation remains almost two-thirds lower over the past 12 months.
Clearly, the commercial challenges it faces required fresh perspectives. Philip Jansen joined as non-executive chair towards the end of 2024, while Cindy Rose was appointed as chief executive in September of this year.
Between them, the two principals collectively acquired £286,900 worth of shares on November 13. Rose recently reassured shareholders that “in [her] first 60 days we are already moving at pace, with some initiatives already announced and more to come”.
Your guide to what Trump’s second term means for Washington, business and the world
The writer is the former chair of the FDIC and author of the upcoming book ‘How Not to Lose a Million Dollars’
Smarting from recent election losses and growing voter discontent over the high cost of living, the Trump administration has renewed its assault on the Federal Reserve for keeping mortgage rates “too high”.
The administration is right to blame the Fed for the high costs of housing, but this is not because it has kept mortgage rates too high. It is because it kept them too low and for too long in years past. Housing affordability continues to be driven by supply shortages. Lower rates will increase demand, making housing inflation worse.
The great financial crisis led to a collapse in home prices and housing construction. In a desperate attempt to revive the economy and property market, the Fed resorted to the controversial practice of quantitative easing to lower longer-term rates on mortgages and US Treasuries.
It purchased vast amounts of mortgage-backed securities, driving up their prices and thus lowering yields. The Fed gradually amassed $1.7tn in MBS — representing about 30 per cent of the total market. Mortgages trended down to the 4 per cent range. Pre-crisis they had hovered about 6 to 7 per cent (about where they are now).
The Fed’s efforts initially made sense. However, it kept QE going far too long. Over the ensuing years demand rose, but supply did not keep pace, creating upward pressure on home prices. Then the pandemic hit. In March 2020, the Fed again embarked on massive purchases of mortgage-backed securities, reaching peak holdings of $2.7tn in 2021. Mortgage rates plunged, sinking as low as 2.65 per cent. Combined with the growing popularity of remote work, this led to a steep increase in the demand for homes, in a still supply-constrained market. The result was average annual home prices spiking by double digits. Home prices increased by 47 per cent between 2020 and 2024.
While ultra-low mortgage rates juiced new demand for housing, it had an even a bigger impact on refinance activity with homeowners rushing to secure lower rates. About one-third of outstanding mortgage debt was refinanced between 2020 and 2021.
But this exacerbated supply constraints. As mortgage rates have normalised, housing inflation has moderated, but it is still too high at about 4 per cent. A major problem is the resale market, which has suffered from chronic undersupply. Homeowners do not want to give up their cheap mortgages by moving. So, the Fed’s actions had a double whammy: they produced red-hot housing inflation while worsening supply imbalances.
This whole sorry history demonstrates the folly of using demand-side stimulus in supply-constrained markets. Yet, the administration continues to prioritise demand-side measures to address housing affordability. It has proposed a 50-year mortgage which, by extending the loan term, would lower monthly payments. This could draw more buyers into housing markets, but only with mortgages that double their interest costs over the standard 30-year mortgage and take decades to build significant equity.
The proposal confuses home affordability with mortgage payment affordability. Working families need homes at prices they can manage and mortgages that help them build wealth.
Blue-state Democrat leadership — which contributed to shortages with restrictive zoning and permitting requirements — has finally caught on that supply is the key problem. California, Colorado, Massachusetts, Maryland and Washington are all states that have embraced more permissive zoning and permitting requirements to expand the supply of low-cost, affordable homes.
Bipartisan legislation, co-sponsored by senators Tim Scott and Elizabeth Warren to cut red tape and provide financial incentives for new housing, has passed the Senate, but it needs a boost from the administration to pass the House. This would be far more productive than Fed bashing.
Fed chair Jay Powell has laudably committed to a continued run-off of the central bank’s portfolio of MBS and vowed that the Fed would not again intervene in that market with purchases.
Will the Fed’s next, Trump-appointed chair succumb to pressure and resume them, though? It would be ironic if, as Democrats finally embrace deregulation to address housing supply, Republicans embrace the failed demand-side approaches that caused the very inflation which helped elect President Donald Trump.
The GOP should own housing affordability as an issue, but to do so it must focus on building more homes, not the illusionary quick fix of monetary policy.
Sam Altman, the co-founder and CEO of OpenAI, the company that developed ChatGPT, recently announced his plans on X to introduce a new adult-only version of ChatGPT.
While the details have not yet been made public, it is speculated that ChatGPT’s erotica will allow verified adult users to generate sexually explicit content, most likely in the form of video, photos, and text-based communication.
ChatGPT has over 800 million active users who engage with the chatbot for everyday tasks such as practical guidance, information seeking, and writing. Besides helping people with these tasks, ChatGPT has transformed how people interact with each other by introducing social companions. These AI generative chatbots can become friends or romantic partners, providing synthetic intimacy for users.
ChatGPT erotica goes one step further than these social chatbot companions by providing a new way to access sexually explicit material. ChatGPT’s erotica is set to launch in December of 2025, according to Altman’s X post.
In a few weeks, we plan to put out a new version of ChatGPT. If you want your ChatGPT to respond in a very human-like way, or use a ton of emoji, or act like a friend, ChatGPT should do it…and we will allow even more, like erotica for verified adults.
This announcement comes right after Altman tightened the restrictions and put up guardrails to protect users after receiving multiple lawsuits claiming that these chatbots contributed to several suicides and acts of self-harm. These restrictions were put into place to protect minors as well as people struggling with mental health disorders; however, it seems that Altman is ready to loosen these restrictions and is now focused on “treating adults like adults.”
“There will still be some controls, of course. No one will see or get erotica unless they ask for it, and the reason OpenAI hasn’t rolled out something like this in the past is concern over people with mental health issues. Now that we have been able to mitigate the serious mental health issues and have new tools, we are going to be able to safely relax the restrictions in most cases,” according to Altman.
Will ChatGPT Lead The Industry In AI Porn?
ChatGPT is not the first AI interface to roll out with AI-generated porn; Elon Musk’s Grok has been the most non-restrictive of the large AI players, allowing users to create AI companions designed for sexual roleplay. Meta has also been in the news after revealing it paid certain celebrities to use their voices in AI companion sexual role play, sexual banter, and romantic fantasies with users.
Without any age restrictions, this may create a dangerous online world for underage users. ChatGPT is the most commonly used chatbot, and launching erotica may open up a whole new world of AI-generated porn to users who know nothing about this concept.
Addressing Concerns About ChatGPT Erotica And Minors
There are many concerns about the rollout of ChatGPT’s erotica. Many critics have expressed their concern, from flaws in the age verification systems, potentially allowing children to access AI porn or become victims of AI-generated images, to the social implications of humans forming intimate relationships with AI chatbots, the potential increase in AI porn addiction, and the adverse effects on mental health. These concerns have been strongly voiced across multiple internet sources.
American business and television personality, Mark Cuban, responded to Altman’s post with major concerns, primarily about the safety of America’s youth. In a response post, he stated, “I’ll say it again. This is not about porn. This is about kids developing ‘relationships’ with an LLM that could take them in any number of very personal directions.”
Cuban believes that the controversy with AI porn and erotica is not about adults, but rather the dangers that erotica can have on underage minors forming emotional relationships with AI without their parents’ knowledge. There has already been a string of heartbreaking suicides involving teenagers and adolescents who developed unhealthy relationships with AI chatbot companions that have resulted in lawsuits against OpenAI. Hence, many think the concern about the relationship between minors’ safety and AI erotica is valid.
There is also concern that children will be able to crack the “age gating”, whether through fake identification or other means. This can be viewed through a similar lens as underage drinking. Even if purchasing alcohol is “age-gated” by means of a government identification, underage drinking still takes place.
Cuban continued on X, “This is going to backfire. Hard. No parent is going to trust that their kids can’t get through your age gating. They will just push their kids to every other LLM. Why take the risk? A few seniors in HS are 18 and decide it would be fun to show the hardcore erotica they created to the 14-year-olds. What could go wrong?”
The Dangers Of Online Age Verification
It is assumed that users will need to scan their government ID to access ChatGPT erotica for age verification and “age gatekeeping”; however, sharing a government ID online could expose users to significant risk of identity theft, especially in the event of a breach of personal information.
This age verification process could make uploading an ID feel routine, but verifying identity online involves third-party verifiers and brokers who link a person’s real identity to everything they do online, including search history, conversations, and even erotica fantasies. This form of surveillance can become a detailed record that cybercriminals can exploit. Critics are concerned that teenagers will find ways to bypass this “age gate” within minutes if they want to, unintentionally revealing their own or someone else’s private information.
Will Human Relationships Take A Back Burner To AI Erotica?
In addition to safety concerns, many people are wary of the effects that AI porn can have on relationships. Questions surrounding relationships and the ethical dilemma of erotica AI, include:
What happens if people are engaging in AI porn without their spouse’s knowledge?
Is it considered cheating?
Will these synthetic relationships with chatbots alter the standards for dating and romance in real life?
Will human romantic relationships fail because a person would rather live out their sexual fantasies with a chatbot?
Will marriages fail?
Will relationships with AI chatbots become an addiction?
While the answers to these questions are yet to be seen, the potential dangers and social implications of how AI porn and erotica could affect human relationships and mental health need to be considered individually before users participate with AI erotica.
ChatGPT’s New Restrictions To Mitigate Mental Health Issues
There has been a lot of conversation from Altman about loosening the restrictions to benefit verified adult users with the upcoming erotica rollout. Still, there was not a lot of talk about the restrictions and tools that were put into place after the string of suicides that were associated with the dangers of AI chatbots.
“We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues. We realize this made it less useful/enjoyable to many users who had no mental health problems, but given the seriousness of the issue, we wanted to get this right,” Altman posted on X.
OpenAI formed an expert council centered around mental well-being and AI. This council consisted of experts in mental health, youth development, and artificial intelligence who collaborated with a network of over 250 physicians, spanning 60 countries, to help develop parental control and safety guidelines, commonly referred to as “guardrails”.
In September 2025, OpenAI introduced a new safety guardrail feature allowing parents to link their ChatGPT accounts with their teens’ (ages 13-17) profiles to allow parents to safeguard their teens’ accounts. This account linking allows parents to:
Adjust AI behavior for age-appropriateness
Disable memory and chat history
Receive alerts if the AI detects signs of acute mental health distress
Block age-inappropriate content such as inappropriate graphics, sexual roleplay, and extreme beauty ideals
Disable image generation to help prevent minors from creating sexually graphic images
Set quiet hours to restrict when their teens can engage with ChatGPT
Opt in to non-personalized feeds and other restrictions in the hopes that this safeguards teens’ mental well-being
OpenAI also programmed ChatGPT to produce healthier, safer responses when user conversations become problematic, including allusions to mental health crises.
Previously, if a user confided in ChatGPT about their mental health struggles, including suicidal ideations and delusions, ChatGPT would sometimes make mistakes and, instead of referring the user to a suicide or mental health hotline, it would offer inappropriate, suggestive advice that would allude to the encouragement of self-harm and suicide, resulting in users taking their own lives. With these updated guidelines, if a user alluded to the fact that they were experiencing delusional beliefs or suicidal tendencies, then the chatbot would offer calming responses that attempted to calm the user and direct them to emergency services.
Another guardrail that was put in place protects users from the overuse of chatbots, which are replacing human relationships. If a user stated that they enjoyed speaking with ChatGPT more than with real people, the chatbot would begin to encourage real-world connections.
AI Erotica’s Role In Porn Addiction
AI erotica not only poses dangers for the mental health and safety of minors but also has the potential to increase the probability of developing a porn addiction. AI porn addiction is an emerging topic in the world of behavioral addictions and mental health. Porn addiction is not yet considered a diagnosable mental health disorder under the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5).
However, porn addiction is considered a type of compulsive sexual behavior disorder (CSBD), which is included in the DSM-5 and is recognized as a mental health disorder. It is characterized by significant distress and impairment due to excessive focus on sexual fantasies, including masturbation, use of pornography, having multiple sex partners, or paying for sex.
Porn addiction can create tension and turmoil in people’s everyday lives, causing financial strain, relationship conflicts, a decline in work performance, and social isolation. Mental health experts believe that porn addiction is deeply rooted in unresolved trauma. Therefore, treatment aims to uncover this trauma and help users develop healthy coping skills that can address this trauma when triggers appear in the future. Additionally, many users compulsively engage in porn as a way to cope with and control their feelings of depression, shame, isolation, or anxiety.
Get Help For Porn And AI Porn Addiction
While the effects of ChatGPT’s erotica rollout are yet to be seen, it is anticipated to increase the risk of porn addiction. If you find yourself struggling with a porn addiction, treatment for porn addiction is available, with therapy often being the first step forward. If you’re ready to get started, explore your online therapy options today.
The contest to stay on the leading edge of AI is rapidly taking center stage in America’s strategic competition with China. But what does it actually mean to beat China in AI? Does the United States have the right strategy for navigating this contest? Are policymakers overstating the threats posed by China, or by artificial intelligence itself? And can Washington and Beijing cooperate on areas of high risk even as they compete intensely elsewhere?
In this episode of Pivotal States, Christopher S. Chivvis speaks with Colin H. Kahl, Director of the Freeman Spogli Institute for International Studies and former Under Secretary of Defense for Policy in the Biden administration, to unpack the peril and the promise of U.S.–China competition in AI.
As above, EUR/USD seems to have survived what could have been a tricky week. Today’s focus will be on the flash PMIs for the region. These have been a source of comfort to the euro in that business sentiment has stayed relatively constructive, suggesting businesses are finding workarounds for the new tariff environment. As an aside, export numbers for November (Korea) and October (Japan) seem to be holding up quite well. A decent set of PMIs today could give the euro a modest lift. We should also see the European Central Bank’s survey of negotiated wages for the third quarter. These are expected at 2.45% quarter-on-quarter annualised versus 3.95% in the previous quarter. This should be welcome news to the ECB, but also a reminder that real wages are rising in the eurozone and, with a high savings rate, eurozone consumption could be a positive surprise for 2026.
Away from the data, we have several central bank speakers today. ECB President Christine Lagarde speaks this morning at the Frankfurt European Banking Congress. The tone of the conference is very much about the benefits of investing in Europe, and we might see references again to the notion of a global euro. Here, we’ve seen reports overnight in Politico that the ECB is thinking of expanding its EUREP repo lines to other central banks outside the euro area. This is an effort to increase comfort in euro invoicing – copying the People’s Bank of China’s playbook for the renminbi – and is something we discussed in our global euro paper earlier this year.
If EUR/USD can somehow make it back above 1.1560/65 today, it will have had a good week.
Elsewhere, we have a Swiss National Bank Watchers conference this afternoon, with plenty of SNB speakers. Expect a similar message along the lines of ‘we have all our options open’ to address Swiss franc strength, when in reality its ability to cut rates and intervene is sorely limited. We see EUR/CHF staying offered near 0.92 into year-end as investors prefer to hold Swiss francs as a hedge against a loss of confidence in government bond markets and fiat currencies.
Only battery or fuel-cell electric vehicles eliminate tailpipe emissions
Flex-fuel cars comprise the majority of the fleet in Brazil, but they cannot deliver the same air quality benefits as electrification. Flex-fuel vehicles have no significant benefits over ICE gasoline vehicles in terms of air pollutants that harm human health. Hybrids may reduce these emissions but don’t eliminate them. In contrast, battery and fuel-cell EVs entirely remove tailpipe pollution from city streets—and all places where Brazilians live, work, and breathe
In addition to their air quality benefits, battery electric cars in Brazil produce substantially lower life-cycle greenhouse gas emissions than flex-fuel, hybrid, or improved ICE vehicles, delivering dual benefits of cleaner air and reduced climate impact. For example, battery electric cars emit about one third of the life-cycle emissions of gasoline-ethanol flex-fuel ICE vehicles.
Yet despite the advantages of battery electric vehicles for both air quality and climate, a recent ICCT study found that the lack of strong EV policies in Brazil has allowed automakers to prioritize flex-fuel and ICE vehicles. This is in stark contrast with other major markets where policies have driven rapid EV deployment, such as China and the United Kingdom, and risks further delaying the health and climate benefits of the EV transition in Brazil.
The choice ahead
The evidence is clear: accelerated EV adoption could deliver billions in health savings through 2040. Behind these figures are children who will breathe easier, families kept whole, and lives extended.
With vehicle electrification, Brazil can secure a win-win for both health and climate. Every electric vehicle replacing an ICE vehicle delivers immediate air quality improvements while cutting greenhouse gas emissions. The global transition to EVs is accelerating. With its clean electricity grid and growing domestic auto industry, Brazil is positioned to lead the way.
Vehicle electrification is underway in Brazil—what remains uncertain is the pace. With automakers increasingly able to supply EVs globally and costs declining steadily, policy choices will determine whether Brazil captures the full health, climate, and economic benefits of this transition, or continues bearing the costs of delaying action.
The environmental non-profit Stand.earth fails to see the wood from the trees when it comes to the Canadian forestry industry and Drax’s limited role within it (Drax still burning 250-year-old trees sourced from forests in Canada, experts say, 9 November). We do not own forests or sawmills, and we do not decide what areas are approved for harvesting.
The vast majority (81%) of our Canadian fibre came from sawdust and other sawmill residues created when sawmills produce wood products used in construction and other industries in 2024. The remaining 19% of our fibre came from forest residues, including low-grade roundwood, tops, branches and bark.
Forests in British Columbia are harvested for lumber by timber companies under strict regulations set by the province’s government in joint decision-making with indigenous First Nations. Around 94% of the province’s forests are on public land, and it is a legal requirement for these sites to be reforested in a free-growing condition. Leaving Canada’s forests unmanaged is not the answer to preserving these landscapes. We witnessed this in Jasper, Alberta last year when a wildfire on unmanaged land led to $880m in damages and significant amounts of CO2 released into the atmosphere.
The fibre highlighted in Stand.earth’s report was low-grade roundwood, which was rejected by sawmills following approved harvesting. Without the biomass sector using this harvest residue, it would likely be controllably burned on site to help mitigate wildfire risk. It is far better to positively use this residual fibre to generate renewable electricity. None of the fibre highlighted in this report came from a designated old growth management area or old growth deferral area. Miguel Veiga-Pestana Chief sustainability officer, Drax Group plc
The UK has a hard-won reputation for global forest protection, thanks to its leadership when it hosted COP in 2021. It is a bitter irony that in the run-up to a climate conference hosted in the Amazon, the UK government has signed a new contract to pay billions in subsidies to Drax (Drax power plant to go on earning ‘over £1m a day’ from burning wood pellets, 5 November). Drax burns millions of tonnes of trees every year, with reporting in this newspaper showing it is still burning trees over 250 years old from Canada.
While claiming to still be a forest leader, the UK spends billions supporting an industry that harms them. This contradiction exposes the government’s hollow claims of environmental leadership. Burning imported biomass is no credible climate solution; it simply shifts emissions and destruction elsewhere.
If Britain truly wants to lead on nature and climate, it must stop financing forest loss under the guise of “green” energy and reclaim the mantle of forest leadership to protect ecosystems. Public money should restore and preserve forests – not reward those who destroy them. Matt Williams Senior forest advocate for NRDC (Natural Resources Defense Council)
This audio is auto-generated. Please let us know if you have feedback.
Andrew Rynhard is chief technology officer for Sidero Labs.
Utilities are increasingly relying on edge computing to support fast, dynamic decision-making across distributed infrastructure. From automated grid balancing to substation control and remote fault detection, edge deployments are helping modernize aging critical infrastructure and drive efficiency.
But as utilities become more software-driven, they’re also introducing new cybersecurity challenges, and nowhere is that more true right now than at the edge.
The expanding edge footprint is a significant security shift. Edge nodes are being deployed at transformer stations, within distributed energy resources, at remote monitoring points and alongside smart meters. These systems often operate in locations where no IT staff is anywhere near, let alone onsite. They may rely on cellular or intermittent connections, and they often run continuously for years without routine maintenance cycles.
For adversaries looking to disrupt utility operations or test the resilience of national infrastructure, edge systems have become an increasingly tempting target. However, the security conversation still tends to revolve around network segmentation, threat detection or endpoint access control. Important as those are, they miss one foundational layer: the operating system.
The overlooked surface: OS-level risk in utilities
Every edge deployment runs an operating system and, in many utility environments, that OS is the weakest link. Traditional Linux distributions (originally built for servers or desktops) still underpin many OT systems, grid controllers and IoT gateways. These OSes are powerful, flexible and familiar. But they weren’t designed for today’s threat environment, nor for the realities of edge deployments that now often run on containerized architecture.
Most conventional OSes are mutable by default. Their configuration can drift, their file systems can be written to by any number of services or processes, and their security settings can be altered over time (often unintentionally). In a centralized data center or enterprise network, these issues are manageable because systems are easy to audit and maintain. At the edge, where access is limited and conditions change, they become liabilities. A system that is secure on Day 1 may no longer be secure on Day 1,000, and you may not know what changed.
For utilities now operating thousands of edge systems, the risk compounds quickly. A small misconfiguration rolled out across 10,000 nodes isn’t just a technical error, but an exploitable pattern. Attackers don’t need zero-day vulnerabilities when they can exploit outdated packages, exposed services or poorly secured update mechanisms.
Why immutability matters at the edge
To meet modern security expectations, utilities’ edge infrastructure needs more than reactive patching or policy enforcement. It also needs to be designed from the ground up to resist tampering, misconfiguration and drift. This is where the concept of an immutable operating system becomes powerful.
An immutable OS is one that cannot be altered during runtime. The system boots into a known-good state (defined and verified ahead of time) and remains in that state throughout operation. No one can log in and manually tweak firewall settings, nor can a rogue process write to the disk. Configuration is declarative, meaning it’s defined through code and automatically enforced on every single node, every single time.
This matters to utilities because the edge is largely inaccessible. If something goes wrong (whether it’s an outage, a breach, or just a silent misconfiguration), physical intervention is costly and slow. Immutable systems reduce the need for human touch. They also make it far easier to reason about security posture at scale. If every node is running the exact same image, with the exact same configuration, verified cryptographically, then audit becomes a matter of validating one system, not thousands.
Immutable systems also simplify updates. Rather than patching live systems in-place (which is a risky prospect in operational technology), you replace the running image with a new, verified version. The update is atomic, meaning it either succeeds completely or fails without altering the running system. That kind of rollback safety is critical when uptime and predictability matter more than raw agility.
As utilities begin adopting containerized architectures at the edge, often orchestrated by lightweight Kubernetes distributions, security becomes even more dependent on a trusted foundation beneath the orchestration layer. If the operating system isn’t secure, then no matter how sophisticated your containers or orchestration tools may be, you’re building on sand.
A real-world shift is underway
Some energy providers are already adopting this model, deploying thousands of edge nodes that balance national grids or support DERs with responsive load adjustment. These projects depend on lightweight, secure operating environments that can be centrally managed, even across geographically scattered deployments.
The early results are promising. Organizations are finding they can gain operational flexibility without expanding their risk surface. They’re reducing overhead by eliminating complex patching cycles and increasing resilience through standardized, locked-down software stacks.
More importantly, they’re setting a new baseline for what “secure by design” means in utilities’ edge deployments. Instead of relying on bolt-on security tools or manual oversight, they’re embedding security into the OS itself.
As the edge becomes a permanent and essential part of utility operations, utilities should look for edge systems that are minimal, declarative and tamper-resistant. The operating system should include only the components required to run its specific workload, avoiding unnecessary packages or exposed services that widen the attack surface.
System behavior should be defined through code (declaratively) and enforced automatically across every deployment. Login interfaces such as SSH should be eliminated altogether, especially in field-deployed infrastructure that cannot be reliably maintained through manual intervention. Security should be verified, not assumed, with cryptographic validation of every system image and configuration at boot. Last but not least, when it comes time to update, the process should be both safe and capable of rolling back cleanly in the event of failure, without leaving the system in an unpredictable state.
For utilities facing expanding risk and rising complexity, these tenets form the basis of a more defensible, manageable edge.
It’s easy to focus security attention on the parts of the stack utilities can see: network traffic, APIs, dashboards, etc. But the OS layer is where everything begins. It’s where services run, where controls are enforced, and where mistakes often take root. As utilities modernize, they should demand the same from their software. Security isn’t just a service, but a system property that must start at the operating system.
