The filing this week of Anthropic’s S-1 ahead of its much-anticipated fall listing teased us with a question we could not resist: how might the company fare in our measurements of corporate stakeholder performance? One might assume that as a public benefit corporation it would do well across the spectrum of criteria we cover. Although the results are highly preliminary, the picture that’s emerging is far from simple.
The positives are clear. A workforce of mostly highly paid engineers, 22 weeks of paid parental leave, a generous 401(k) match, a new employee stock purchase plan, and education assistance would place it near the top on pay and benefits. The company creates good jobs – headcount rose from about 2,300 in December 2025 to more than 5,300 by September 2026 – and has no history of widespread layoffs. It also discloses on many of the key AI governance practices we know matter to the public: a commitment against mass surveillance, human oversight of its models, and board-level oversight of AI. Only 14% of the 1,002 largest US companies disclose a human-oversight policy, and just 10% address harm, deception or manipulation.
Oh, and philanthropically it could become one of the most generous companies in American history, with 5% of equity pledged to a public benefit fund (worth $75bn-100bn on projected valuations), and founders pledging 80% of their wealth.
However, the elephant in the room is obvious. A company chartered “for the long-term benefit of humanity” states in more than 90 pages of disclosed risks that its own product “could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity.”
In recent years we’ve seen companies pull back from overtly involving themselves in social issues. In the age of AI, business and society will become more intertwined than ever. Frontier labs could be the most just companies ever, helping cure diseases, unlock entrepreneurship, solve the climate crisis and uplift the human condition. Or they could undermine – through labor market disruptions, cyber security risks, or worse – the very security and social stability on which healthy markets and democracy depend. Of the pieces I’ve read recently on the subject, Jordi Visser’s latest substack (The Price of Abundance) captures it perfectly; “The great paradox of AI is that economic productivity could accelerate dramatically while the traditional pathways through which individuals achieve prosperity become less reliable.”
Business as usual is over. The traditional measures of corporate impact on society need an upgrade. Government, business and civil society leaders need to work together to find solutions. We are up for the challenge.
Be well,
Martin
64% of the public believe companies should put more than 5% of AI investments toward AI safety. Explore the data here.
Donna Morris, Executive Vice President & Chief People Officer at Walmart, writes on LinkedIn about the learnings the company has had integrating AI into their workforce, and how training makes all the difference.
Axios explains how “Black October” is going to reshape the future of cybersecurity, with renewed emphasis on how quickly a company can detect an intruder, contain the damage, and recover.
Microsoft is offering buyouts to nearly 60,000 employees (27% of its workforce) to manage the costs of its AI build-out. Reuters has the story.
Bloomberg reports on how a bipartisan bill to remake US permitting laws and expedite the development of energy projects promises to shave electric bills and increase real wages nationwide per the White House Council of Economic Advisors.
CBS News looks at what private sector jobs have seen large jumps in pay over the last several years.
The New York Times looks at how the current lawsuit against Dollar General is emblematic of Americans losing faith that corporations are pricing their products fairly.
Axios questions whether a surge in American productivity is a result of AI or other business factors.
