AI-discovered PHP flaws prompt patch rush and government site outages
PHP's developers said "critical vulnerabilities" were found in the language, and websites were taken offline on Wednesday and Thursday. The flaws carry a severity score of 9.2 out of 10 and would let an unauthenticated attacker run code remotely on servers. Fixes were released in versions 8.5.12, 8.4.27, 8.3.36 and 8.2.35. According to Wiz, PHP powers 70% of the world's websites.[29][30]
The Conversation says researchers from Wiz used Google's Gemini and OpenAI's ChatGPT models to trawl PHP's codebase in August. It says PHP developers confirmed the issue by late September, and that a Hacker News user leaked news of the flaws before patches were ready, so developers confirmed the issue publicly on October 21 and released emergency patches early.[28]
Amazon said it identified an attempt to exploit the bug four days before public disclosure and that the flaws are now "broadly exploited in the wild". The BBC reports no successful attacks so far, and the NCSC's Olly Whitehouse said there were "no indications" of successful breaches. Security firms report attempts without confirmed success.[27][29][30]
UK minister Baroness Lloyd told the House of Lords that 32 government services were unavailable and four partially unavailable. The Netherlands took portals including DigiD offline as a precaution, and Orange suspended some services globally. The Australian Signals Directorate, US CISA and bodies from the UK, Europe and Japan urged urgent updates, and the EU's computer emergency response team advised isolating exposed applications if patching is not possible.[28][29][30]
Where outlets differ. Amazon says the flaws are being broadly exploited, while the BBC, citing the NCSC, reports no indications of successful breaches and MercoPress reports attack attempts without confirmed success.
[27][29][30] Microsoft says its AI outperforms doctors on hard diagnostic cases
The findings were published Thursday in Science. Microsoft claims its tool can diagnose conditions up to four times more accurately than human doctors. Microsoft AI CEO Mustafa Suleyman told The Verge "it is finally here".[32]
TechCrunch reports that in a simulation using New England Journal of Medicine cases, the system was correct 83% of the time versus 36% for primary care doctors and general internists working individually with reference materials. Specialists and physician teams, who usually handle such cases, were not compared. Gains did not appear on emergency department cases, and the study's data suggests much of the improvement comes from more powerful underlying models.[31]
OpenAI brings its frontier models to Google Cloud
Under an agreement announced Wednesday, Google Cloud customers will be able to buy OpenAI's frontier models through Google's Model Garden and Marketplace, including with existing cloud spending commitments. The models will also come to Google's Antigravity environment, with general availability planned in the coming weeks.[33][34]
Ars Technica reports OpenAI says Microsoft remains its primary cloud partner. Axios says the deal follows OpenAI's expansion onto Amazon Web Services earlier this year. Analyst Mike Leone told Axios that once every cloud has top models, the advantage goes to whoever does best with a company's data, governance and costs.[33][34]
Canva releases first image generation model, Canva Craft
Canva launched Canva Craft, which uses a two-stage architecture that plans a design before generating an image, producing up to 2K resolution in more than 30 aspect ratios. Canva says this allows "greater precision, consistency and control".[35][36]
TechCrunch reports the model was trained on human-designed templates from Canva's library rather than photographs. It is available to all users via the Dream Lab tab and the Canva API, and Canva said two cheaper versions will be released soon.[35]
AI firms post record lobbying quarters
Federal disclosures show Anthropic, OpenAI and Nvidia each had their highest-spending lobbying quarter between July and September. Nvidia spent $2.2 million, Anthropic $2.1 million and OpenAI $1.5 million. Meta spent $5.9 million, the most of any big tech firm in the period, according to Semafor.[37][38]
Anthropic's form lists a bill allowing labs to coordinate on security risks; OpenAI's lists "AI-security collaboration and information-sharing, including proposed antitrust protections". Semafor notes the forms do not say whether companies support or oppose proposals. The Verge notes Microsoft has spent at least $2 million a quarter since 2021 and Google regularly over $3 million.[37][38]