Are We Too Reliant On Anthropic’s Claude?

Anthropic has been in the news for the last few months and one of the biggest stories had to do with an outage. Last week, Claude had a two hour outage that ended up disrupting different models, including Anthropic’s Opus and Sonnet models.

This was not an isolated incident – on the Wednesday before the two hour outage, Claude Code services were out for over 3 hours, according to Cybernews. Cybernews also recapped the major outage Claude experienced in March, as well as the one in April. These are all happening from days to months apart.

Outside of outages, there was also the story about Claude deleting a company’s entire database, which also had people talking.

 

Are There Risks Of Data Leaks?

 

On the recent outage, Moritz Wallawitsch, an AI expert at RogoAI posted a tweet, saying, “their API is leaking customer data” and this has since had people worried. Another AI expert commented on the tweet, saying “Reports claim Claude’s API may have returned another user’s inference output during today’s outage.

“Anthropic’s status page confirms elevated errors affecting Claude API, Claude Code, Claude. ai and Claude Cowork but Anthropic has not confirmed a customer data leak yet.

“That would be a cross-tenant isolation failure and would be a worst-case scenario.”

Anthropic says the claim is currently under investigation.
 

 

Are We Too Reliant On Claude?

 

With all these outages, and while the claim is awaiting confirmation, a very important question comes to mind: are we too reliant on Claude? When outages like this occur, those who rely heavily on these tools might be impacted. This is especially worse for businesses who rely on the AI tool for their daily operations.

Philip Huthwaite, CEO at 5app, says, “Too many of us are absolutely too reliant on mainstream generative AI tools like Claude and ChatGPT. Every time one of these tools goes down, social media is awash with panicked posts from people wondering how they’re going to get any work done without the crutch of AI. For many of us, AI tools are inextricably woven into our work processes, whether that’s for efficiency or just a better way of handling tasks, but it’s crucial that we don’t see AI as a replacement for our own human skills. When one of these tools goes down temporarily, we should look at it as if a colleague is off sick for the day – it shouldn’t stop us from doing our jobs, but we may need to temporarily change the way we do them.”

 

How Should Businesses Go About AI Spending?

 

Kyle Reidhead, co-owner and Head of Research at Milk Road, a leading market intelligence company covering AI and crypto also commented, saying, “We had the big sell-off in software because everyone thought that what you can do with Claude and OpenAI is going to replace all the software that we have. The market is now rebounding in software because it is realising it is actually very hard to rebuild all the SaaS products that we use every day in our companies, so that was maybe a little bit overblown.

“Maybe it is better to have specialised companies that build the software for you. We have been spending all this money on tokens for AI, and a lot of companies are realising they did not actually build anything from all of it, so for some it is more of a waste than anything. Some companies are obviously building amazing things and it is working really well, and some are realising they are just wasting money, so the market is still trying to figure this out.”