The SharePoint attacks might be an inside job

Welcome to Runtime! Today: Cybersecurity experts are worried about a program designed to help companies respond to vulnerabilities following the SharePoint attacks, Microsoft and OpenAI are reportedly making progress on a new deal, and the latest funding rounds in enterprise tech.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

They don't love you like I love you

Microsoft's enterprise software tools helped countless companies and government organizations around the world take their first steps onto the information superhighway back around the turn of the century. Keeping that enormous installed base of software safe as new vulnerabilities are discovered requires Microsoft to enlist security partners around the world to help quickly distribute patches when things go wrong, but what if those partners can't be trusted?

Security experts are starting to wonder if there is a mole in the Microsoft Active Protections Program (MAPP), which gives security companies "early access to vulnerability information so that they can provide updated protections to customers faster." Bloomberg reported Friday that Microsoft is concerned that someone in the MAPP program leaked details about SharePoint vulnerabilities to hackers working on behalf of the Chinese government, who attacked vulnerable SharePoint servers "the day before it released a patch publicly," according to the report.

• After going through a vetting process and signing a non-disclosure agreement, MAPP members receive patches for Microsoft vulnerabilities a day earlier than they are released to the public, which allows those partners to help their own customers respond more quickly when a vulnerability is disclosed.
• One huge benefit of cloud computing is that vendors can patch vulnerabilities on behalf of their users before they are publicly released, but companies that manage their own servers are on their own to install the patches.
• Security partners can help those companies respond much faster when new vulnerabilities are being actively exploited.

But the process didn't work this time around for several reasons, and at least five U.S. government agencies were hit by attacks exploiting the vulnerability. Some of the blame can be attributed to Microsoft's incomplete patch for the initial flaw, but only program insiders would have known how to get around it that quickly. 

• Dustin Childs of Trend Micro laid out a timeline of events for The Register: The initial flaw was discovered and responsibly disclosed as part of a hacking contest earlier this year, and Microsoft prepared a set of patches for the July edition of Patch Tuesday, which was July 8th.
• However, vulnerable SharePoint servers found themselves under attack from groups Microsoft said were working for the Chinese government on July 7th, when MAPP members would have received details on the patches.
• The first set of patches didn't close the loophole all the way, which didn't become clear for a couple of weeks until Microsoft acknowledged that SharePoint servers were under active attack on July 19th.
• By that point, hundreds of companies had been hit.
  

"At least a dozen Chinese companies participate in the [MAPP] initiative," according to Bloomberg. Security experts have been concerned for many years about a law passed in 2021 that requires Chinese security researchers to report newly discovered flaws to the government before disclosing them to the public.

• Leaks in the MAPP program have happened before, but this month's incident could prompt new changes that restrict the flow of information.
• "If I thought a leak came from this channel, I would not be telling that channel anything," Childs told The Register.
• But given the immense complexity of the modern software supply chain, coordination is essential to helping protect companies themselves from attacks.
• And if vibe coding tools produce an explosion in the amount of software created on a regular basis that many see coming, managing the explosion in vulnerabilities in that software — without trusted partners — will become even harder.

Maybe AGI was the money we made along the way

The Microsoft-OpenAI contract renegotiations continue to be quite leaky considering the billions of dollars at stake if OpenAI can't find a way to restructure itself as a for-profit company. This week's episode comes courtesy of Bloomberg, which reported that the two companies are close to figuring out a way to allow Microsoft continued access to OpenAI's technology.

Under the terms of their current agreement, which runs through 2030, OpenAI can end its partnership with Microsoft once it has reached "AGI," as nebulous a term as any thrown around during the generative AI boom. There is no industry-wide agreed-upon definition of what constitutes "AGI," but it "can be triggered by technical or business milestones" as specified in their current agreement, according to Bloomberg.

That means Microsoft could be left high and dry if OpenAI declared that GPT-5, which is expected to arrive next month, is the pinnacle of AI achievement. As should be expected, the particulars will come down to money: OpenAI wants a larger share of the revenue Microsoft gets from selling access to its models on Azure, and Microsoft wants a sizable stake in the for-profit version of OpenAI.

Enterprise funding

BlinkOps raised $50 million in Series B funding for its no-code agent-building tool, which helps customers automate security practices.

Dropzone AI landed $37 million in Series B funding as it builds agents for security operations teams that promise to help them sort through alerts and issues.

Julius scored $10 million in seed funding for its natural-language data-analysis tool.

Nebulock launched with $8.5 million in new funding as it tries to modernize endpoint security with AI.

FluidCloud launched with $8.1 million in seed funding for its multicloud infrastructure management tool.

The Runtime roundup

Anthropic introduced new rate limits for some Claude customers, and according to VentureBeat software developers were not exactly thrilled by that move.

Freshworks beat Wall Street estimates for revenue and profit and raised its annual revenue guidance, citing demand for its AI services.

Thanks for reading — see you Thursday!