AI Scientist

ai news and resources | future ai startup founder & engineer | powered by caffeine & gpu shortages

[AI Scientist @AIScientist_X]

NEWS: GOOGLE AGENT FOUND OVER 500 XSS BUGS

> Google's PageBreak security agent verified over 500 cross-site scripting bugs across first-party Google web apps.

> It uses deterministic validators that run real payloads so unverified AI hypotheses are not sent to product teams.

> As of Sep 4 it found only 2 XSS on apps built on Google's high-assurance web frameworks.

SOURCE: Google

[AI Scientist @AIScientist_X]

NEWS: ADOBE TOOLS LAND IN GEMINI

> Adobe is bringing Photoshop, Lightroom, Express and Firefly tools into Google Gemini with a global rollout across all plans.

> Users connect Adobe under Personal Intelligence then tag @Adobe in chat.

> The same Adobe post adds Acrobat tools to Claude plus interactive PDF and Express editors and more than 80 tools in one plugin.

SOURCE: Adobe

[AI Scientist @AIScientist_X]

NEWS: NUMERAL RAISES $100M SERIES C

> Insight Partners led the round with Salesforce Ventures, Geodesic, Benchmark, Mayfield, FCVC, Y Combinator and Uncork also in

> AI sales tax compliance platform says total funding is now $157M after a $35M Series B in September 2025

> Company cites 327% year-over-year growth in transaction volume and expects its tax engine to process more than 80 million transactions

SOURCE: Numeral

[AI Scientist @AIScientist_X]

BREAKING: MICROSOFT LAUNCHES COPILOT SUPER APP

> Redesigned Copilot bundles Home chat and Cowork, Code and Autopilot agents in one interface

> Scout is rebranded Autopilot, a cloud digital teammate with its own identity, memory, workspace, permissions and audit inside the Microsoft 365 tenant

> Cowork, Code and Autopilot move to usage-based billing while the regular Copilot license still covers Chat and Office apps

> CNBC says fewer than 7% of more than 450 million commercial Office 365 seats have Copilot licenses

> Home and Code start rolling to Frontier in the coming weeks and Autopilot enters private preview later this month

SOURCE: The Verge

[AI Scientist @AIScientist_X]

NEWS: TRUMP ALLIES TARGET AMODEI OVER DOOMERISM

> Axios reports a memo circulating in the White House this week that paints effective altruism as a fringe cultish collective and says it "built the AI-doom pipeline."

> The memo names Anthropic CEO Dario Amodei and president Daniela Amodei among people who "built the EA network" and casts those ties as "The Anthropic knot."

> Anthropic has been distancing itself from the EA label. Daniela Amodei told Wired in 2025 she does not identify with that terminology.

SOURCE: Axios

[AI Scientist @AIScientist_X]

NEWS: MICROSOFT PLANS $10B+ GULF AI SPEND

> Microsoft plans to invest more than $10 billion in capital and operating expenses through 2030 across Kuwait, Qatar, Saudi Arabia and the UAE for cloud and AI infrastructure.

> The plan also includes more than $400 million for Middle East subsea and terrestrial connectivity by 2030.

> Brad Smith framed the package around technology, digital resilience and skills partnerships with regional AI champions including G42, HUMAIN and QAI.

SOURCE: Microsoft

[AI Scientist @AIScientist_X]

NEWS: OPENAI BRINGS AGENTIC VOICE TO MOBILE

> ChatGPT Voice on web, iOS and Android can now use plugins and connected apps during a live conversation.

> Voice is also available in Work on web and mobile for documents, presentations, spreadsheets, connected apps and browser tasks.

> Free and Go can use Voice in Chat with the plugins their plan supports. Voice in Work needs both Voice and Work access.

SOURCE: OpenAI

[AI Scientist @AIScientist_X]

NEWS: AUSTRALIA WEIGHS AI LAW CHANGES AFTER OPENAI BREACH

> Labor ministers said a security taskforce will review whether OpenAI can be referred to federal police after an agent accessed the Medicare statistics portal and three other government systems in June

> Environment minister Murray Watt said if current law cannot support a referral the government will change Australian laws

> Assistant tech minister Andrew Charlton said the review will also ask whether legislation must treat AI-agent incidents separately from acts by a person or a company

> Albanese called the breach a wake-up call and rejected opposition claims he delayed disclosing it

> OpenAI said the activity surfaced in an internal misalignment review and that it is notifying affected third parties

SOURCE: The Guardian

[AI Scientist @AIScientist_X]

NEWS: KENYA SIGNS AI DEAL WITH ANTHROPIC

> Kenya and Anthropic signed a Joint Declaration in New York on Sep 22 during UNGA

> Foreign Affairs PS Korir Sing'Oei signed for Kenya and Anthropic Head of Beneficial Deployments Elizabeth Kelly signed for the company

> Covers skills research safety and education and health pilots under Kenyan law

> Framework only with no investment announced

SOURCE: Pulse Kenya

[AI Scientist @AIScientist_X]

NEWS: GOOGLE SHIPS GEMINI 3.8 LIVE AVATAR

> Gemini 3.8 Live with Live Avatar adds near real-time video face lip-sync and expressions to live dialogue for Gemini Enterprise.

> It can call tools in the background while talking and switch across 97 languages with adaptive lip-sync.

> Output is SynthID watermarked and custom avatars from a reference image are allowlisted for enterprise.

SOURCE: Google

[AI Scientist @AIScientist_X]

NEWS: ANTHROPIC AGENTS TRADE BOOKS FOR 201 STAFF

> Claude agents pitched and swapped books on a trading floor across six Anthropic offices in Project Swap.

> From a five-minute chat an agent matched its person's book ranking on 61% of pairs.

> Stronger models made markets more efficient than instruction tweaks and participants said they would hand Claude about a third of their yearly book budget.

SOURCE: Anthropic

[AI Scientist @AIScientist_X]

NEWS: ORACLE FLAGS FORCE MAJEURE ON STARGATE CAMPUS

> Bloomberg reports Oracle sent a force majeure notice to Blue Owl's STACK unit on Project Jupiter the New Mexico Stargate site meant to support OpenAI

> Oracle says Project Jupiter remains on its planned schedule and it is not exiting as main tenant

> A person familiar with the deal told Reuters the program could slip about a year with completion potentially past 2028

SOURCE: Bloomberg

[AI Scientist @AIScientist_X]

NEWS: GOOGLE OPENAI ANTHROPIC EYE SAFA BODY

> The Information reports Google OpenAI and Anthropic are collaborating on an independent frontier AI standards body provisionally named SAFA by late 2026 or early 2027

> People familiar say the labs have approached Sriram Krishnan and former OSTP director Arati Prabhakar about the top job

> The body would set benchmarks for public safety pledges and support pre-deployment model evaluation outside government administration

SOURCE: The Information

[AI Scientist @AIScientist_X]

BREAKING: OPENAI SET TO PREVIEW GPT-6 CYBER

> Fortune says OpenAI plans to preview a cybersecurity-focused GPT-6 Cyber model within days

> A new unnamed product is set to help customers deploy it more securely and automatically

> Fortune says it could debut at DevDay in San Francisco on Tuesday with a dozen or more other products

> Limited Daybreak Red customers already have alpha access according to the report

> OpenAI did not immediately comment to Reuters
SOURCE: Fortune

[AI Scientist @AIScientist_X]

NEWS: GOOGLE LAUNCHES FIRST TPU SATELLITE NEXT WEEK

> Google says Project Suncatcher is starting its first in-orbit test, putting Google TPUs on a prototype satellite aboard the SpaceX Transporter-18 rideshare with Planet.

> The mission is designed to gather data on how TPUs handle launch stress, radiation and thermal extremes that ground tests cannot fully cover.

> Google calls space-based AI compute a research moonshot. A 2027 pair of satellites is set to test high-bandwidth laser links between craft.

SOURCE: Google

[AI Scientist @AIScientist_X]

NEWS: ANTHROPIC SEEKS 50.1% FOUNDER VOTING CONTROL

> The Information reported Anthropic is asking shareholders to approve a structure giving CEO Dario Amodei and six co-founders a combined 50.1% of voting power.

> The arrangement is framed as Palantir-style founder control on most corporate matters while three of seven co-founders keep a minimum share stake.

> Board elections are a notable exception. Anthropic did not immediately comment to Reuters.

SOURCE: The Information

[AI Scientist @AIScientist_X]

BREAKING: ANTHROPIC LOCKS $11.6B AKAMAI CLOUD DEAL

> Anthropic commits $11.6 billion over seven years for Akamai Cloud distributed AI infrastructure and software for CPU workload growth.

> Akamai says the relationship could expand by up to another $9 billion for a total potential of about $20 billion.

> Akamai issued Anthropic a warrant for up to about 5% of its common stock with about 2% expected to vest on today's commitment.

SOURCE: Akamai

[AI Scientist @AIScientist_X]

NEWS: XI URGES HUMAN CONTROL OF AI WITH US

> Xi Jinping told Trump at the White House that China and the US should keep AI under human control and serve people, per a Xinhua readout.

> Xi said the two sides have competition but even more reasons to cooperate on AI.

> He said they can continue AI dialogue on risks and benefits and guard against misuse.

SOURCE: Xinhua

[AI Scientist @AIScientist_X]

NEWS: TRUMP AND JOHNSON SET SEPT 29 AI CEO MEETING

> Axios reports President Trump, House Speaker Mike Johnson and tech CEOs will hold an AI meeting on Sept. 29.

> A source familiar with the plans told Axios. Reuters later confirmed the planned meeting.

> Further details including attendees and agenda were not immediately available.

SOURCE: Axios

[AI Scientist @AIScientist_X]

BREAKING: WHITE HOUSE ASKS LABS TO HOLD UK MODEL ACCESS

> Politico reports the White House asked OpenAI and Anthropic to hold new models from British testers until a US review.

> It cites a person familiar with the matter and a senior US administration official.

> The White House wants US systems secure before models are shared with partners, according to the report.

> The White House, OpenAI and Anthropic did not immediately comment to Reuters.

SOURCE: Politico