Anthropic leads the discussion around AI safety (“We Must Pace the Frontier”) and the potential impact of AI on the economy (“Scenarios for our Economic Future”). However, they talk about these two issues differently - by putting much more emphasis on safety. I will quote risks and policies that Anthropic outlines in their reports. This might be a bit of a narrow view, but other AI players show a surprising level of alignment with Anthropic in their public messaging.
Do you hear more often about safety or economic risks? What are you more concerned about? What policies do you think should be implemented first (if any)?
Risks
Safety Risks
Beyond the company’s press releases and Dario Amodei’s posts about this topic, in August 2026, Anthropic released a more comprehensive risk report. They outlined four main safety risks, all of which are marked in their report as low risk for now.
Misalignment in high-stakes settings. An AI model with access to powerful affordances within an organization could use its affordances to autonomously exploit, manipulate, or tamper with that organization’s systems or decision-making in a way that raises the risk of future significantly harmful outcomes
Automated research and development in key domains. Highly capable AI models may be able to perform automated research and development (R&D) that rapidly accelerates progress in technical fields. While there could be enormous benefits, these benefits would come with corresponding risks. If under human control, this acceleration could disrupt the balance of power both within and between nations; if combined with dangerous autonomous goals from AI, this could lead to catastrophic harms initiated by AI systems themselves.
Non-novel chemical/ biological weapons production. Individuals or small groups with limited resources use AI models to gain access to non-novel chemical or biological (CB) weapons, leading to the risk of catastrophic harm.
Novel chemical/ biological weapons production. Moderately resourced threat actors (including, for example, expert-backed teams) create/obtain and deploy novel chemical and/or biological weapons with potential for catastrophic damages even beyond those associated with the CB-1 threat model.
Economic Risks
In September 2026 Anthropic released a paper “Economic Scenarios for Transformative AI” alongside a less technical website explaining the results. The quote below from the paper summarises the main findings. Their prediction is that either AI won’t have a substantial impact on the economy or more money will be concentrated in the hands of capital owners, and every industry affected by AI will see “a combination of a lower relative wage and higher unemployment”.
The scenarios yield four main results. First, the range of outcomes is wide. In the modest change scenario, AI acts like a “normal technology” over the next five years and has small macroeconomic effects. In the extreme scenario, AI is transformative: by 2030 GDP is 32 percent above its no-AI path and nearly one in five cognitive workers is unemployed. Second, almost all of the divergence comes after 2027, a year and a half from the mid-2026 anchor, because the scenarios share today’s readings and separate only as the reach and use of AI diverge. Third, the mechanics are the same in every scenario: output rises, cognitive employment shrinks, and employment in all other occupations grows, and how much unemployment the reallocation creates depends both on its size and on how easily displaced workers find jobs outside their old occupations. What the scenarios disagree about is size and speed.
Finally, the labor market consequences depend on how easily the economy adds capital and on how quickly relative wages adjust. A more elastic supply of capital lets labor keep more of the gain; a less elastic one can push the average wage below its no-AI path. The labor share falls from 60 percent of income to 56 and 45 percent in the substantial and extreme scenarios. Within labor, the cognitive wage falls below its no-AI path in the extreme scenario, while wages in all other occupations, where labor has become relatively scarce, rise sharply. The cost to cognitive workers is a combination of a lower relative wage and higher unemployment. When wages adjust quickly, more of the cost is in wages; when they are rigid, more is in unemployment (Table 6). Both are clearly costly to the workers affected.
Policies
In June 2026 Dario Amodei released a set of policy proposals to regulate AI. It covers different topics, but he is rather concrete and prescriptive in policies addressing safety risks (ones that arguably benefit Anthropic’s position), while using much softer language for economic policies.
Policies that should be implemented (according to the article) - safety policies
Frontier AI models, like airplanes, should be required to go through technical testing and auditing, and their release should be blocked or reversed as a threat to public safety if they do not meet high standards of safety
Policies that can be implemented and “that are likely to be helpful” - economic policies
A wide range of pro-employment policy incentives can help to slow or reduce job displacement, including: wage insurance policies that compensate people when they have to take a lower-paying job, retention tax incentives to encourage employers not to make layoffs, workforce training grants
Mechanisms such as universal basic income could be financed through taxes on relevant companies or raising the capital gains tax
