This week, the Pentagon gave Anthropic an ultimatum: if it removed ethical restrictions on AI, it would lose its $200 million contract and be placed on a government “blacklist.” Separately, also this week, Anthropic relaxed its basic security policy to give the company more freedom to grow in a competitive and dynamic market.
It’s unclear how this week will play out for Anthropic’s business and reputation, but its decisions will have consequences.We know this because Anthropic’s dilemma is familiar in the tech industry. Many companies boast about their values and morality only to face difficult decisions that force them to choose between growth and upholding those ideals.
OpenAI and the Altman Weekend
In one of the strangest board dramas in corporate history, Anthropic’s main rival, OpenAI, abruptly fired its founder and CEO, Sam Altman, on a Friday in November 2023, only to rehire him the following Tuesday.
The saga involved a unique corporate structure that placed the rapidly growing, for-profit company behind ChatGPT under the auspices of a nonprofit board of directors. Four years earlier, the company had written in its bylaws that OpenAI remained “concerned” about AI’s potential to “bring rapid change” to humanity. The company’s overseers feared that Altman would act so quickly that he risked undermining the security the company had pledged to provide.
But Altman’s dismissal triggered threats of a mass exodus of employees, an unsustainable situation that could have bankrupted the company. As a result, the board rehired Altman a few days later. The board dissolved shortly thereafter, and Altman restructured the company last year to free itself from its nonprofit oversight body.
Since then, OpenAI has struggled to find a balance between speed and security, and has faced several lawsuits alleging that its products convinced young people to self-harm. OpenAI denies these allegations.
Apple and the San Bernardino attacker
Investigators obtained permission to retrieve data from Farook’s iPhone, but were unable to access it because it was locked with a passcode. A California judge ordered Apple to assist law enforcement in accessing the phone.
But in an open letter signed by Apple CEO Tim Cook, the company refused. Cook stated that the judge’s order would open a backdoor into the iPhone, something they considered too dangerous to create. The company stated it had no sympathy for terrorists, but that complying with the order would grant government authorities the power to access anyone’s device and capture their data.
Apple faced heavy criticism for its decision, including from then-presidential candidate Donald Trump. However, it has since received widespread praise for upholding customer privacy, which has become a hallmark of the company’s brand.
The company now routinely promotes that it will not sell customer data or store certain personal information on its servers, trying to differentiate itself from Google, one of its main competitors.
Etsy vs. individual sellers
When Amazon’s e-commerce empire was just beginning to take off in the early 2000s, Etsy emerged as a novel alternative where shoppers could find unique, handmade products.
But in 2013, it implemented a controversial change that threatened to undermine that philosophy. It expanded its policy to allow sellers to source from manufacturers and outsource operations, which at the time raised concerns that it would no longer provide a level playing field for small, independent sellers without the resources to hire staff.
Even so, that decision was fundamental for Etsy to expand into the marketplace it is today, offering more than 100 million items for sale and boasting approximately 8 million active sellers.
“From a business standpoint, it worked for Etsy, but it was a difficult time for the company,” said Arun Sundararajan, director of the Fubon Center for Technology, Business and Innovation at NYU Stern.
What’s next for Anthropic?
Now, Anthropic’s biggest short-term challenge is likely whether customers and potential customers value and trust the company, said Owen Daniels, associate director of analytics at Georgetown’s Center for Security and Emerging Technology.
Anthropic stated that its self-imposed security measures were always flexible and subject to change as AI evolves. It pledged to be transparent about security going forward and stated that it really had no other choice: if it stopped growing, competitors who didn’t value security as much could gain ground and make AI “less secure” overall.
It is unclear what will result from Anthropic’s change, because the existential risks of AI are still largely “conceptual,” Sundararajan noted.
He said he would be wary of any expert who considered this a pivotal moment for AI security. But it could be for the company.
“For me, the fact that Anthropic has backtracked on a specific security promise has more to do with Anthropic than with the future of AI,” he said.