The Next Web · Aug 3
The White House finalized its frontier-model review framework, then declined to publish what is in it
After the August 1 executive-order deadline, the administration walked AI companies through a finalized framework that gives the government up to a 30-day pre-release review window to assess the cybersecurity capabilities of frontier models before they ship. The framework is unclassified but is not being made public, and the benchmarks used to judge a model plus the capability threshold that triggers review are treated as classified. One official said unclassified does not mean they will broadcast it, and the NSA is tied to the classified benchmarks.
▸ The MSP Angle
Is the government about to control which AI models can ship?
The rules that decide which frontier models reach the market, and how fast, are now being written in a document the public cannot read, which turns your upstream AI supply into a variable you cannot fully forecast. A 30-day review window means the next capable model your stack depends on could arrive on a schedule set by classified benchmarks rather than a vendor roadmap. You cannot influence this, so build for it: keep your platform model-agnostic so a delay or gate on any one vendor is a rerouting decision, not a stalled client project.
Read at The Next Web ↗
EU AI Act · Aug 2
The EU can now fine AI model vendors up to 3% of global revenue as GPAI enforcement goes live
As of August 2, the European Commission's supervision and enforcement powers over general-purpose AI model providers came into force. The AI Office can now compel documentation, evaluate models directly, order corrective measures, and restrict or withdraw a model from the EU market. Under Article 101, fines reach 3 percent of a provider's total worldwide annual turnover or 15 million euros, whichever is higher. The obligations themselves had applied since August 2025, but the Commission could not enforce them until now.
▸ The MSP Angle
Can EU regulators actually fine the AI vendors my clients use?
Yes, starting now, and the enforcement target is the model provider, not just the deployer, which changes the platform-risk math for any client you serve in or into the EU. A vendor that can be fined or pulled from the EU market is a continuity risk you should be able to route around, so know which providers sit behind each client workload and whether you can switch if one gets restricted. This is the enforceable phase the earlier timeline debate was leading to: treat the compliance calendar as real and make model portability part of every EU-facing client's plan.
Read at EU AI Act ↗
The Hacker News · Aug 5
CISA flags an actively exploited RCE in the Langflow AI agent builder
On August 5 CISA added three actively exploited flaws to its catalog, led by CVE-2026-9198, a code injection bug in Langflow that gives unauthenticated attackers full remote code execution on default deployments. Langflow is a popular low-code framework for building AI agents. The same batch added an Apache Tomcat flaw whose exploitation was tied to the DeepSeek-driven autonomous attack campaign reported last week, plus the original N-central bypass.
▸ The MSP Angle
Are the tools my clients use to build AI agents a security risk?
The low-code agent builders showing up in client environments are now in the actively-exploited column, and a default Langflow install exposed to the network is unauthenticated remote code execution waiting to happen. Shadow AI is not only chatbots, it is the agent-building tools developers stand up without telling anyone. Add these frameworks to your external scanning and asset inventory, because the same client who asks you to secure their firewall may have an agent builder listening on an open port that you have never been told about.
Read at The Hacker News ↗
Anthropic · Aug 4
Anthropic names a former state supreme court justice as its first Chief Global Affairs Officer
Anthropic announced on August 4 that Mariano-Florentino Cuellar, a former California Supreme Court justice and former president of the Carnegie Endowment for International Peace, will be its first Chief Global Affairs Officer, leading policy, international engagement, and government relationships worldwide. He has held roles across three presidential administrations and teaches at Stanford Law School. The hire lands as the labs and Washington negotiate how frontier models get reviewed.
▸ The MSP Angle
Why is a top AI lab hiring a former judge to run policy?
A frontier lab staffing up with a former supreme court justice is a signal that AI regulation is becoming a defining business fight, not a side conversation, and the outcome shapes the tools you put in front of clients. Model vendors are now organizing to influence rules on model access, safety testing, and liability that will flow downstream into your vendor contracts. You do not need to follow the politics, but you should expect the terms of service and compliance obligations on the AI you deploy to keep shifting as this plays out.
Read at Anthropic ↗
Salesforce · Jul 24
Salesforce lands a $1.6 billion agentic AI deal with the VA as Agentforce revenue passes $1.2 billion
The Department of Veterans Affairs signed a three-year, 1.6 billion dollar agentic enterprise license agreement with Salesforce covering Agentforce Public Sector and Agentforce Health, aimed at triage, intake, and care coordination across a virtual contact center that has already handled more than 40 million veteran calls. Salesforce separately reported Agentforce annual recurring revenue of about 1.2 billion dollars, up 205 percent year over year.
▸ The MSP Angle
Is agentic AI actually generating revenue yet, or is it hype?
A nine-figure government agent deal and a billion-dollar recurring-revenue line are proof the agentic market is real money, not a demo, which is the validation to bring into client conversations that stall on whether agents are ready. The tell for your practice is where the value landed: a contact center handling millions of calls, meaning high-volume, repetitive service work is the agentic sweet spot. Point clients at their own highest-volume, most repetitive workflows first, because that is where agents pay for themselves fastest.
Read at Salesforce ↗
ChannelE2E · Jul 28
An operator-backed platform is buying up MSPs to bolt AI services onto mid-market shops
e4n, an operator-led platform founded by executives experienced in scaling tech-services firms, backed its portfolio company Katalyst in acquiring Layer27, a Durham managed services and cybersecurity provider, in Katalyst's first tuck-in since the May 2026 partnership. The stated model preserves acquired teams and brands while layering shared operations, group purchasing, and AI-driven automation, adding healthcare-sector experience and AI readiness to Katalyst's base.
▸ The MSP Angle
Are private equity rollups turning AI into an MSP land grab?
Capital is now buying MSPs specifically to add AI services and operational scale, which raises both the valuation ceiling for well-run independents and the competitive pressure on shops without an AI story. An acquirer that markets AI readiness as the reason to buy is telling you what the market now prices, whether you plan to sell or stay independent. Build a real, differentiated AI practice now: it is the asset that either commands the premium at exit or defends your clients from a better-capitalized competitor.
Read at ChannelE2E ↗
Challenger, Gray & Christmas · Aug 1
AI led US job-cut reasons for a fifth straight month, even as hiring rose
The July report from Challenger, Gray and Christmas counted 33,429 announced US job cuts, the lowest monthly total in two years, with AI named as the leading reason for the fifth consecutive month at 10,970 cuts. Technology led all sectors at 149,023 cuts year to date, up 67 percent over the same period in 2025. Hiring announcements rose 25 percent year over year, which the firm says shows AI is shifting the labor market rather than dismantling it.
▸ The MSP Angle
Is AI actually causing layoffs, or is that overblown?
AI has been the top stated reason for job cuts for five months running, but hiring is up at the same time, which is the nuance to give clients who are either panicking or dismissing it. The pattern is a reshuffle, not a collapse: work is moving, not disappearing. That is the exact moment small and mid-size clients look for a partner to help them redeploy people toward higher-value work instead of cutting blindly, and an MSP that frames AI as capacity rather than replacement wins the trust that a tools-only vendor cannot.
Read at Challenger, Gray & Christmas ↗
TechCrunch · Aug 3
A Benioff-backed startup raised $20 million on the premise that someone has to make enterprise AI actually work
June, founded by the team behind an AI company Salesforce acquired in 2019, raised a 20 million dollar pre-seed round led by Marc Benioff's Time Ventures with Michael Dell, Aaron Levie, and George Kurtz backing it. Its pitch centers on forward-deployed engineers, specialists who work inside a company to scan its systems, find bottlenecks, and stand up agent-powered processes, on the premise that before AI creates value, someone has to deal with fragmented data, complex workflows, and years of technical debt.
▸ The MSP Angle
Why do so many companies struggle to get real value from AI?
Investors just put money behind the exact problem MSPs already solve: enterprise AI does not work until someone untangles the legacy systems, the fragmented data, and the technical debt underneath it. The forward-deployed engineer this startup is selling is a role you can already staff, embedded in the client environment, mapping the workflows before deploying an agent. Sell the integration work, not the model. The client who cannot get AI to stick does not need another tool, they need the person who makes the tools fit their business, and that person can be you.
Read at TechCrunch ↗
Google Workspace Updates · Aug 3
Gemini in Classroom expands to students of all ages, on by default, starting August 10
Google is extending Gemini in Classroom to K-12 and higher-education students of all ages who already have access, with a web rollout starting August 10 and mobile on August 17. Access to the Gemini tab is on by default for teachers and students where Gemini in Classroom is enabled, and is governed by the Gemini in Classroom admin control. If the control is off for students, it stays off, including for those under 18.
▸ The MSP Angle
What changes for education clients when Gemini reaches all students?
For any MSP administering Workspace for a school, this is an obligation with a date on it: a generative capability turns on by default for students of all ages starting August 10, so the governance decision has to be made before then, not after a parent asks. Confirm the Gemini in Classroom control matches each client's data-handling and age policy now, and document the choice. The default-on setting means silence is a decision, and it is better to make it deliberately than to discover it in a support ticket.
Read at Google Workspace Updates ↗
