The Hacker News · Sep 7
A perfect-10 N-central flaw lands as Huntress probes a patched server's breach
N-able shipped hotfixes for a maximum-severity pre-authentication remote-code-execution flaw in N-central (CVE-2026-86218, CVSS 10.0) alongside two authentication bypasses. Huntress separately investigated a fully patched N-central production environment compromised on September 4, and thin appliance logging meant it could not confirm which exploit the intruder used.
▸ The MSP Angle
My RMM is patched, so am I safe?
Patched is the floor, not the finish line: the compromised server Huntress examined was current on updates, and the appliance kept too little history to reconstruct the break-in. Apply the N-central hotfixes today, then fix the part patching cannot: forward RMM logs somewhere an attacker cannot erase them and alert on new admin accounts. Your management plane deserves the same monitoring you sell.
Read at The Hacker News ↗
VentureBeat · Sep 1
Anthropic's Claude Fable 5.1 keeps its price but cuts cached-input costs 75%
Anthropic released Claude Fable 5.1 on September 1 at unchanged token rates while cutting cache-read pricing from $1.00 to $0.25 per million tokens, projecting bills 25% to 45% lower on agent-heavy workloads. Benchmark gains were steep, including 55.8% on Terminal-Bench 4.0, and a restricted Mythos 5.1 variant ships only to vetted cybersecurity and life-sciences organizations.
▸ The MSP Angle
Why did my AI costs change when the sticker price did not?
Agent workloads reread the same instructions, documents, and tool definitions constantly, so cached-input price is where the real money moves. If you bill AI services at a fixed monthly rate, a 75% cache cut quietly improves your economics, and knowing that before your competitor does is an edge. Track effective cost per workload, not the list price on the vendor's page.
Read at VentureBeat ↗
NVIDIA · Sep 3
Nvidia is buying Hugging Face, the home of open AI models, for $12.9 billion
Nvidia agreed to acquire Hugging Face for about $12.9 billion, roughly $11.9 billion in cash plus up to $1 billion in retention equity, its second-largest deal ever. The platform hosts more than 3 million models and 500,000 datasets for 18 million developers, and Nvidia says it stays open, multi-cloud, and free of any requirement to run on Nvidia compute.
▸ The MSP Angle
Does it matter to my clients who owns the big AI platforms?
Consolidation is reaching the open layer of the AI stack, and acquisition-day promises have a way of eroding by year three. Nothing changes for clients today, but note which of your tools and vendors depend on Hugging Face for models or hosting. A portable posture, open weights you can rehost and workloads that can move, is what keeps a client from inheriting someone else's lock-in.
Read at NVIDIA ↗
GlobeNewswire · Aug 27
GTIA: SMBs are going slow and steady on AI, and half already use a partner
GTIA's survey of 520 SMB and microbusiness decision-makers found 40% deploying AI against specific business needs, 84% reporting positive impact, and half already working with a strategic partner on AI initiatives. The top asks map directly to services: employee training (38%), picking the right tools (38%), integration (31%), and security or compliance review (29%).
▸ The MSP Angle
What AI help are small businesses actually willing to buy?
The support gaps SMBs name are all services an MSP can staff this quarter: training, tool selection, integration, and security review. Half of these buyers already route AI work through a partner, so the deals are going to whoever shows up with a packaged offer first. Sell the groundwork, policies and training, then expand into the automation work it uncovers.
Read at GlobeNewswire ↗
Challenger, Gray & Christmas · Sep 2
AI fell to fourth among layoff reasons in August after five months on top
Challenger, Gray & Christmas counted 52,881 announced U.S. job cuts in August, the lowest August total since 2022, with restructuring the top stated reason and AI down to fourth at 3,462 cuts. AI still leads for the year at 116,175 cuts, about 22% of all announcements, while tech-sector cuts are up 52% year over year.
▸ The MSP Angle
Is the AI layoff wave getting worse or leveling off?
One quiet month is not a reversal, but it gives clients a more accurate picture than the headlines: AI-driven cuts concentrate around rushed, ungoverned rollouts, and the firms investing most heavily have been hiring. Use the data to steer clients toward augmentation with training rather than quiet replacement. That version of AI adoption is also the one an MSP can package, deliver, and defend.
Read at Challenger, Gray & Christmas ↗
NPR · Aug 18
Grads blame AI for a brutal job hunt; economists say the data is messier
NPR profiled graduates sending hundreds of applications into a market where recent-grad unemployment hit 5.7% versus 4.1% overall, and 47% told ZipRecruiter that AI has affected hiring in their field. Stanford's Erik Brynjolfsson finds a 16% relative employment drop for young workers in AI-exposed roles, while Harvard's David Deming notes junior hiring fell before ChatGPT launched and points at remote work instead.
▸ The MSP Angle
Should I stop hiring junior staff because AI does the entry-level work?
The research is genuinely split, so firms making permanent staffing decisions off one scary chart are guessing. A Ramp study of 21,000 firms found the heaviest AI adopters grew entry-level hiring 12%, which suggests the winning pattern pairs juniors with AI rather than skipping them. For your own bench and your clients', hire people who work well with agents and automate the tasks nobody should be hired for.
Read at NPR ↗
Techstrong.ai · Aug 18
Google hands its Agent2Agent protocol to a neutral foundation
Google contributed A2A, the protocol that lets AI agents from different vendors find each other and hand off work, to the Linux Foundation's Agentic AI Foundation, placing it alongside the Model Context Protocol under neutral governance. More than 150 organizations back the standard, including Google, Anthropic, OpenAI, and AWS, and researchers are already warning that agents trusting each other's output blindly is the next security problem.
▸ The MSP Angle
Will AI agents from different vendors work together for my clients?
Neutral governance for agent protocols is good news for anyone who refuses to bet a client on one vendor's ecosystem, because interoperability is what makes automations portable. The caution sits inside the same story: an agent chain treats upstream output as trusted input, so one bad or compromised agent can poison everything downstream. Interoperability plus agent-level access control is the architecture to ask every platform about.
Read at Techstrong.ai ↗
Transparency Coalition · Sep 4
California just sent about 30 AI bills to the governor's desk
California closed its 2026 session on August 31 with roughly 30 AI bills passed, and Governor Newsom has until September 30 to sign or veto them. The stack includes AB 1883, which bars workplace surveillance tools that read workers' neural data or emotional state, chatbot disclosure and teen-safety rules, and health-care AI confidentiality requirements, while New York and New Jersey enacted their own AI laws this summer.
▸ The MSP Angle
Do state AI laws affect my clients if we are not in California?
California rules have a habit of becoming everyone's rules, the way CCPA reset privacy expectations nationally. The recurring themes, disclosure when AI is talking, limits on AI monitoring of employees, and human override in high-stakes decisions, are cheap to adopt early and expensive to retrofit. Build them into your standard AI deployments now and compliance becomes a selling point instead of a scramble.
Read at Transparency Coalition ↗
Tech Startups · Sep 8
Cognition closed $2 billion at a $48 billion value as Devin revenue doubled
Cognition raised a $2 billion Series E at a $48 billion valuation, up from $26 billion in May, led by Andreessen Horowitz and Accel. Run-rate revenue for its Devin software-engineering agent climbed from $492 million to nearly $900 million in four months, and investors priced the round off that measured traction rather than a demo.
▸ The MSP Angle
What separates an AI product worth paying for from hype?
The market is rewarding agents with measurable output, and Devin's revenue doubled because customers can count finished work against cost. Apply the same test to every AI service you sell: name the metric the client will see move, then report on it monthly. Agents that finish tasks earn renewals; assistants that only suggest things earn shrugs.
Read at Tech Startups ↗
CNBC · Sep 3