Microsoft’s latest outage on Tuesday might have been amplified by its own unforced errors, the company said in an incident report.

“While the initial trigger event was a distributed denial-of-service (DDoS) attack, which activated our DDoS protection mechanisms, initial investigations suggest that an error in the implementation of our defenses amplified the impact of the attack rather than mitigating it,” the report said.

The Microsoft 365 outage on Tuesday is the latest in a series of unforced errors by major IT vendors.

Failure to adequately test systems before roll-out was also a factor in the CrowdStrike incident on July 19, and behind DigiCert’s short-notice revocation of erroneously issued SSL certificates earlier this week.

The July 19 incident was caused by a flaw in CrowdStrike’s security sensor software that cost users millions of dollars in repairs and lost business opportunities, and that testing had failed to uncover.

A root cause analysis of the DigiCert incident showed that there were some process failures during the modernization of a software system that had also been missed during testing.

Steps Microsoft took to mitigate the outage

The latest problems with Microsoft 365 began to appear around 11:45 UTC on Tuesday, when an unexpected usage spike resulted in Azure Front Door (AFD) and Azure Content Delivery Network (CDN) components performing below acceptable thresholds, leading to intermittent errors, timeout, and latency spikes, Microsoft said. 

The dip in performance affected a subset of Microsoft 365 services and other services, including Azure App Services, Application Insights, Azure IoT Central, Azure Log Search Alerts, Azure Policy, as well as the Azure portal itself.

The services impacted included the Microsoft 365 admin center itself, Intune, Entra, and Power Platform.

In response to the outage, the company said that it had started investigations immediately and once it understood that a DDoS attack was behind the network spike, it had implemented networking configuration changes to support its DDoS protection efforts and performed failovers to alternate networking paths to provide relief.

“Our initial network configuration changes successfully mitigated majority of the impact by 14:10 UTC,” the company wrote in the report.

However, it pointed out that despite its early efforts several enterprise customers complained of less than 100% availability, which the company began mitigating at 18:00 UTC.

Without giving further details in the incident report, Microsoft said that it used a different approach to try and solve the issue starting with Asia Pacific and Europe.

“After validating that this revised approach successfully eliminated the side effect impacts of the initial mitigation, we rolled it out to regions in the Americas. Failure rates returned to pre-incident levels by 19:43 UTC,” the company wrote in the incident report, adding the incident was finally mitigated at 19:43 UTC.

Additional steps promised by Microsoft

In its initial report Microsoft said internal teams will be completing an investigation to understand the entire incident in more detail.

“We will publish a Preliminary Post Incident Review (PIR) within approximately 72 hours, to share more details on what happened and how we responded. After our internal retrospective is completed, generally within 14 days, we will publish a Final Post Incident Review with any additional details and learnings,” the company wrote in the report.

This is Microsoft’s 8th service status-related incident this year, according to the company’s service status page.

Last year was also riddled with outages for Microsoft 365 users. Azure’s service page shows that the last incident reported in 2023 was in September, when the US East region faced issues.

Reddit, which has signed cooperation agreements with Google and Open AI — giving both companies the right to train their AI models using the site’s content — is now demanding that Microsoft, Anthropic, and Perplexity do the right thing and sign similar agreements.

According to Reddit CEO Steve Huffman, the three companies have repeatedly used Reddit content to train their AI models, despite not being allowed to do so. Reddit has tried to block access via an updated version of the robots.txt file, but that hasn’t stopped the targeted companies from continuing to collect data.

A spokesperson from Anthropic said in a comment to The Verge that the collection of data from Reddit stopped in mid-May. Microsoft and Perplexity, however, did not immediately commented on Huffman’s claims.

Google parent Alphabet’s $2 billion investment in AI firm Anthropic has caught the eye of the UK’s antitrust regulator.

On Tuesday, the UK’s Competition and Markets Authority (CMA) opened an inquiry into whether Alphabet’s partnership with Anthropic created a “relevant merger situation” that threatened competition within the fast-growing market for cloud-delivered AI products and services.

Invitation to comment

The CMA is inviting industry comments before a deadline of Tuesday, August 13, in advance of the launch of its formal investigation. The outcome of the inquiry will decide whether or not the regulator orders remedial actions or otherwise intervenes in the market.

The prospective probe is part of more general concerns about competition in the generative AI industry.

Partnerships between other hyperscalers and AI startups have also attracted concerns, including the business relationships between Microsoft and OpenAI as well as links between Amazon Web Services and Anthropic.

Open relationship

Google told Computerworld that Anthropic is free to partner with other cloud technology providers and hyperscalers, effectively arguing that competitive concerns were misplaced.

“Google is committed to building the most open and innovative AI ecosystem in the world,” the tech giant said in a statement. “Anthropic is free to use multiple cloud providers and does, and we don’t demand exclusive tech rights.”

In a statement, Anthropic told Computerworld it intends to “cooperate with the CMA and provide them with the complete picture about Google’s investment and our commercial collaboration.

“We are an independent company and none of our strategic partnerships or investor relationships diminish the independence of our corporate governance or our freedom to partner with others,” the company said. “Anthropic’s independence is a core attribute, integral both to our public benefit mission and to serving our customers wherever and however they prefer to access Claude.”

Smaller players in the cloud computing market argue that powerful partnerships threaten the development of competition in the AI marketplace.

‘Virtual monopoly’

Josh Mesout, chief innovation officer of UK-based cloud computing firm Civo, told Computerworld, “As an industry we should be cautious over powerful partnerships as they pose a threat to the entire ecosystem by suffocating competition and innovation.”

He added, “We cannot surrender AI to a virtual monopoly before it has really started.”

Maintaining a diverse and competitive landscape in artificial intelligence is important not least because of the diverse and far-reaching applications of AI technologies across multiple industry sectors.

“Over-dependence on a handful of major firms could stifle innovation, limit consumer choice, and potentially lead to a monopoly that favors Big Tech,” Mesout warned.

“To keep the market fair and open, regulators should be eyeing these types of partnerships warily,” he said. “Otherwise, we risk AI following the path of cloud, where hyperscalers run unchecked and leave a broken, locked-in, and stifled market in their wake.”

To get a sense of what a smarter Siri in iOS 18.1 might look like once it appears, Open AI just introduced a new voice mode in its app, albeit in limited alpha, meaning not every user will get ahold of the new tech.

Delayed for a month in response to quality concerns, this is a test of the company’s Advanced Voice Mode on ChatGPT; it’s available to iPhone users who subscribe to the $20 per month ChatGPT Plus service using its GPT-4o model. 

The company warns that it might make mistakes and says access and rate limits are subject to change. It isn’t expected to be universally available across all users until the end of the year, and should be available to Mac, iPhone, and iPad users once it appears. Subscribers accepted to the alpha group will get an alert in the app and an email inviting them to take part in the test. 

“We’ll continue to add more people on a rolling basis and plan for everyone on Plus to have access in the fall,” OpenAI said.

What does Advanced Voice Mode do? 

Effectively, it’s a more powerful chatbot that delivers more natural, real-time conversations with a degree of contextual awareness, which means it can understand and respond to emotion and non-verbal cues. It is also capable of processing prompts more swiftly, which significantly reduces the latency within conversations, and lets you interrupt it to get it to change what it says at any time.

OpenAI first demonstrated the new mode in April, when it showed how the tool can recognize different languages simultaneously and translate them in real time. During that demo, employees were able to interrupt ChatGPT, get it to tell stories in different ways, and more. One thing the bot can no longer do is sound like Scarlet Johansson — it now supports only four preset voices in order to prevent it being used for impersonation. OpenAI has also put filters in place to block requests to generate music or other copyrighted audio, reflecting legal challenges raised against song-generating AI firms such as Suno.

Video and screen sharing capabilities are not yet available.

How it works

If you are a ChatGPT Plus subscriber running the latest version of the app, and are accepted to the test, you can access the bot from within the app by tapping the Voice icon at the bottom of the screen. You can then switch between the new Advanced mode and the existing Standard mode (better for longer sessions) using an interface at the top of the screen. Privacy concerns mean many Apple users might prefer to access these features via Apple Intelligence.

What about privacy?

Apple Intelligence puts additional safeguards in place to protect people’s privacy. As Wired points out, ChatGPT’s user agreement at present appears to want to use your voice and images for training purposes. In a remarkably quotable line, AI consultant Angus Allan calls it a “data hoover on steroids. Their privacy policy explicitly states they collect all user input and reserve the right to train their models on this,” he said.

This is less a problem when used with Apple Intelligence, as ChatGPT requests are anonymized and data from those sessions is not used to train ChatGPT models, according to Apple. If that proves true, many Apple users will eventually gravitate to accessing ChatGPT via their Apple AI as the safest way to use it.

All eyes now will turn to Google, which is expected to introduce similar features within Google Gemini AI soon — features that might also end up being integrated inside Apple Intelligence. The battle of the bots is heating up.

Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.

The demand for AI platform software is expected to grow 40% a year over the next four years, rising from $27.9 billion in sales last year to $153 billion in 2028, according to a new report from research firm IDC.

The report focused on the rapid pace by which AI platforms, such as Microsoft Azure AI, Amazon AI services, Google Cloud AI, and OpenAI grew last year, and how that growth is projected to maintain a “remarkable momentum,” driven by the increasing adoption of technology across many industries.

IDC expects that level of growth to push revenue for AI software to $307 billion worldwide in 2027. That forecast includes platforms and AI applications, AI System Infrastructure Software (SIS), and AI Application Development and Deployment (AD&D) software.

In 2023, the global AI platforms market grew by 44.4% year-on-year compared to 2022. Microsoft led the market, increasing by 77.9% last year to capture 13.8% of the market. Palantir, a major AI player, had 7.5% of the market, representing an 18.2% year-over-year increase, according to IDC.

“OpenAI’s meteoric rise in 2023 marked nothing short of an enormous transformation in the AI landscape,” IDC said in its report: “Worldwide Artificial Intelligence Platforms Software Market Shares.” OpenAI had a staggering 690% year-over-year increase in revenue last year; the company’s market share soared to 5.8%, “a remarkable achievement for a relative newcomer in this highly competitive field,” IDC said.

 Ritu Jyoti, IDC’s group vice president of IDC’s AI, Automation and Analytics research, said the current market shows “no signs of slowing down. Rapid innovations in generative AI is changing how companies think about their products, how they develop and deploy AI applications, and how they leverage technology themselves for reinventing their business models and competitive positioning.”

AI platform adoption will continue to accelerate with the emergence of unified platforms for predictive and generative AI that support interoperating APIs, ecosystem extensibility, and responsible AI adoption at scale, according to Jyoti.

IDC expects cloud-based deployments of AI software to grow at a faster rate than on-premises deployments, with revenue from AI platforms in the public cloud forecast to have a five-year CAGR of 50.9%.

“This trend is attributed to the advanced security measures, data and regulatory compliance, and the scalability capabilities that cloud vendors offer,” IDC said. “With the rapid advancement of technology and the growing demand for AI solutions from businesses across industries, cloud-based deployment of AI platforms software is expected to continue expanding at a rapid rate.”

A leading internet standards body has slammed Google for deciding once again to leave in place support for third-party cookies in its Chrome browser — even though the privacy-invading tracking technology “is not good for the Web” and can harm users.

Google’s announcement last week that it still isn’t dropping support for third-party cookies came “out of the blue” and “undermines a lot of the work we’ve done together to make the Web work without third-party cookies,” Hadley Berman, of the Worldwide Web Consortium (W3C), wrote in a blog post Monday. (The post was titled: “Third-party cookies have got to go.”)

The W3C agrees with the updated RFC definition of cookies, which acknowledges they have “inherent privacy issues.” Moreover, the RFC strongly recommends that “user agents adopt a policy for third-party cookies that is as restrictive as compatibility constraints permit.”

While third-party cookies — which are set by a website other than the one a user is visiting through embedded content such as ads, social media widgets, or tracking pixels — can be helpful when used for authentication across multiple sites, they also enable hidden data collection about users’ internet activity, Berman said. 

There also are other hazards lurking in “the tracking and subsequent data collection and brokerage” that third-party cookies support, including “micro-targeting of political messages” that harm society at large, she wrote.

Google’s ‘user’s choice’ approach to cookies

Rather than end support for third-party cookies, Google instead decided to update Chrome’s cross-site tracking protection policy, unveiled last December, with an option in the settings of Chrome’s Privacy Sandbox, a set of privacy-preserving APIs. The option allows users to choose whether they want to experience web browsing within the Privacy Sandbox setting or continue to have traditional cross-site cookies activated. 

Chrome users can also use the “Enhanced Ad Privacy” feature Google rolled out last year as part of Chrome version 115; it allows for interest-based advertising without tracking individual users across websites, the company said.

The W3C has been working with Google’s Privacy Sandbox team for several years on third-party cookie policies with “substantial progress,” Berman noted. The recent change in direction by Google represents a major step back in that effort, she said.

“The unfortunate climb-down will also have secondary effects, as it is likely to delay cross-browser work on effective alternatives to third-party cookies,” Berman wrote. “We fear it will have an overall detrimental impact on the cause of improving privacy on the web.”

That said, the W3C hopes Google “reverses this decision and re-commits to a path towards removal of third-party cookies,” she added. 

Google did not immediately respond to requests for comment Tuesday.

Google’s lack of privacy leadership

Privacy experts acknowledged that while third-party cookies do present privacy concerns, there are numerous stakeholders to consider.

“Google has repeatedly attempted to replace cookies…aiming to balance user privacy with the needs of advertisers,” said Jason Soroko, senior vice president of product at Sectigo, a provider of certificate lifecycle management. “However, these efforts have struggled due to resistance from privacy advocates, regulatory hurdles, and technical challenges.”

That likely contributed to Google’s decision to delay pulling its support for cookies, he said, citing the “complex interplay between innovation, privacy concerns, and regulatory frameworks.”

More disappointing is that the company “still seemingly has no clear plan to implement greater privacy and safety controls against tracking,” said one privacy expert, who doesn’t believe Google is doing enough.

Google “has long boasted about the innovation happening in its Privacy Sandbox initiative, but that has yet to publicly bear fruit,” said Gal Ringel, Cofounder and CEO at Mine, a global data privacy-management firm.

Moreover, given Google’s role as “the single most influential organization on the internet today,” the company’s failure “to take a true stand on privacy sets a bad precedent on the issue at a critical time when the US is trying to pass more legislation to address the problem,” he added.