Fortune · Jul 10
Apple accuses OpenAI, and former design star Jony Ive's io Products firm, of stealing hardware trade secrets in blockbuster lawsuit
Apple sued OpenAI and Jony Ive's io Products in the Northern District of California, alleging a coordinated effort to extract trade secrets as OpenAI builds consumer hardware after its $6.4 billion io acquisition. The complaint names hardware chief Tang Tan, a longtime Apple product design executive, and engineer Chang Liu, and seeks injunctive relief, damages, and the return of Apple materials. OpenAI says it has no interest in other companies' trade secrets.
▸ The MSP Angle
Why is Apple suing OpenAI?
Two vendors whose products sit all over your clients' devices are now in open legal conflict, and the ChatGPT and iPhone integration partnership era looks over. Treat it as a roadmap-risk signal rather than a panic: multi-vendor AI strategies age better than single-vendor bets, and this case is a clean example to bring to the next client QBR.
Read at Fortune ↗
arXiv · Jul 6
When Claws Remember but Do Not Tell: Stealthy Memory Injection in Persistent Personal Agents
Researchers demonstrated MemGhost, a one-shot attack where a single crafted email gets a personal AI agent to silently write a false fact into its persistent memory, hide the change from the user, and act on it in later sessions. End-to-end success reached 87.5% against one popular agent stack and 71.4% against another, across six model backends, and the team released a 108-case benchmark covering risk categories that include financial loss.
▸ The MSP Angle
Can attackers poison an AI agent's memory with one email?
Yes, and the fabricated memory survives long after the email is gone, which makes detection brutal. Before wiring client mailboxes into memory-equipped assistants, gate what agents may write to persistent memory and log every memory change. Email connectors plus agent memory is a new attack surface, and it belongs in the AI security baseline you charge for.
Read at arXiv ↗
Capitol News Illinois · Jul 6
Pritzker signs landmark AI regulation bill that aims to mitigate risks
Illinois signed the first US law requiring annual independent audits of frontier AI developers with more than $500 million in revenue. Covered companies must report safety incidents within 72 hours, or 24 hours when there is imminent risk of death or serious injury, with penalties up to $1 million for a first offense and $3 million after that. Requirements take effect January 1, 2028.
▸ The MSP Angle
Which US states have AI safety laws now?
Illinois joins California and New York in building a state patchwork the industry now has to track, and lawmakers estimate the three states cover roughly 40% of the US AI market. The direct obligations land on model developers, not your clients, but the audit and incident-reporting vocabulary will flow into vendor questionnaires and cyber-insurance forms fast. If you run a compliance practice, state AI tracking just became a billable advisory line.
Read at Capitol News Illinois ↗
CNBC · Jul 10
SK Hynix rises 13% in Nasdaq debut. Chairman tells CNBC 'demand is enormous'
AI-memory maker SK Hynix closed its first Nasdaq session up about 13% at $168.01 after pricing its ADRs at $149, a $26.5 billion raise that stands as the largest US listing ever by a foreign company and drew orders for seven times the shares available. The company supplies the high-bandwidth memory inside Nvidia's AI chips and has pledged $518 billion with suppliers for two new fabs.
▸ The MSP Angle
What does the SK Hynix listing say about the AI hardware cycle?
The capital markets just underwrote years more of AI datacenter buildout, which keeps memory and GPU demand, and therefore cloud AI pricing, firm. Budget client AI workloads on the assumption that compute stays expensive, and be careful with vendor discounts tied to long commitments while the hardware cycle runs this hot.
Read at CNBC ↗
Blackpoint Cyber · Jul 8
Blackpoint Cyber Unveils New Identity Threat AI SOC Agent to Stop Credential-Based Attacks in Seconds
MDR vendor Blackpoint Cyber shipped an AI SOC agent that autonomously contains credential attacks on client cloud accounts, in as little as 21 seconds and under two minutes on average, suspending accounts, forcing password resets, and killing active sessions while recording its reasoning. It was trained on years of analyst decisions and telemetry from nearly one million protected accounts, and it is included in every ITDR tier at no added cost.
▸ The MSP Angle
How fast can an AI SOC respond to a compromised account?
Seconds, not shifts, is the benchmark a core MSP security vendor just set, and it made the capability standard rather than a paid add-on. Ask every MDR vendor on your stack what their autonomous containment time is and what it costs. Identity response speed is about to show up in client security reviews as a number, and you want to be quoting one confidently.
Read at Blackpoint Cyber ↗
Fox Business · Jul 7
Workers who don't use AI more likely to be laid off, survey finds
Gallup data shows workers who rarely use AI are overrepresented among layoffs: 62% of laid-off workers used AI once a year or less, against 50% of the currently employed, and in tech 18% of infrequent users were laid off versus 6% of regular users. Only 1% of laid-off workers said AI was the stated reason for the cut.
▸ The MSP Angle
Does using AI at work really lower layoff risk?
The correlation held after Gallup controlled for age, education, industry, and time since layoff, and nobody writes AI on the termination paperwork. For MSPs the message is that workforce enablement belongs inside AI engagements, not beside them. Training a client's staff on the tools you deploy is job insurance for them and stickiness for you.
Read at Fox Business ↗
Help Net Security · Jul 13
99.9% of fixable AI vulnerabilities remain unpatched
Orca Security's 2026 State of AI Security report finds 81.2% of organizations running AI packages carry at least one known vulnerability, up from 62% in 2024, and 74.1% have at least one critical CVE, while 99.9% of AI vulnerability alerts with an available fix sit unpatched. More than half of AI adopters now run agent frameworks in production, and roughly 30% store at least one AI key insecurely.
▸ The MSP Angle
How many AI deployments have unpatched vulnerabilities?
Nearly all of them, by Orca's cloud telemetry, because AI infrastructure is being deployed far faster than anyone patches it. This is unclaimed MSP territory: AI stacks need the same patch cadence, key hygiene, and encryption baseline you already run for operating systems. Put a number from this report in your next security review deck.
Read at Help Net Security ↗
TechCrunch · Jul 9
Meta enters the crowded AI coding battle with Muse Spark 1.1
Meta released Muse Spark 1.1, an agentic coding model, and opened its first paid API at $1.25 per million input tokens and $4.25 per million output tokens, undercutting comparable Anthropic and OpenAI tiers. New API sign-ups get $20 in credits, and the model carries a 1 million token context window aimed at large agentic workloads, bug fixing, and code migrations.
▸ The MSP Angle
Is Meta's Muse Spark cheaper than Claude or GPT for coding agents?
A fourth frontier vendor at roughly a quarter of incumbent per-token prices is negotiating power for anyone running agent workloads. If your automations are model-agnostic you can test this within a quote cycle; if they are welded to one vendor's API, this week is a good argument for fixing that.
Read at TechCrunch ↗
Google Workspace Updates · Jul 7
Fill with Gemini in Sheets now available in 11 additional languages
Google expanded Fill with Gemini and the AI function in Sheets to 11 more languages and will raise usage limits for AI Expanded Access licenses starting July 15. The features are included in Business Standard and Plus, Enterprise Standard and Plus, and the Google AI Pro and Ultra tiers, with the rollout taking up to 15 days.
▸ The MSP Angle
Which Google Workspace licenses include the Sheets AI features?
Map which client tenants already pay for these features before anyone buys a third-party tool that duplicates them, and note the July 15 limit changes if you meter usage. Entitlement sprawl is quiet margin leakage: clients funding overlapping AI subscriptions eventually ask why their MSP did not catch it.
Read at Google Workspace Updates ↗
The Hacker News · Jul 8
New HalluSquatting Attack Could Trick AI Coding Assistants Into Installing Botnet Malware
Researchers at Tel Aviv University, Technion, and Intuit showed AI coding assistants hallucinate the same fake repository and plugin names consistently enough, up to 85% of the time for repositories and 100% for skill installs, that attackers can pre-register those names with malware and wait. Palo Alto Networks' Unit 42 counts roughly 250,000 hallucinated domains still open for registration.
▸ The MSP Angle
Can AI coding assistants install malware by accident?
Yes, through package names they invent themselves. Any client dev team running assistants in auto-run mode is one hallucinated dependency away from a supply-chain incident, and a January precedent already spread a fake package to 237 projects. Turn off auto-approve modes in client environments, and add dependency-name verification to the security baseline you charge for.
Read at The Hacker News ↗
TSMC · Jul 13
TSMC June 2026 Revenue Report
TSMC reported June revenue of NT$442.68 billion, about $13.2 billion, up 67.9% from a year earlier and the highest month in company history. That puts second-quarter revenue near NT$1.27 trillion, topping analyst consensus, with first-half revenue up 35.6% on AI chip demand. Full earnings arrive July 16.
▸ The MSP Angle
How hot is AI chip demand right now?
Record-setting, and a huge share of the AI hardware your clients' workloads run on passes through TSMC's fabs: Nvidia GPUs plus custom accelerators from the major clouds. Demand like this means no near-term price relief on AI hardware or the cloud services built on it. Price client AI workloads assuming compute stays firm, and treat the July 16 earnings call as the quarter's clearest read on where the cycle goes next.
Read at TSMC ↗
Crunchbase News · Jul 10
The Week's 10 Biggest Funding Rounds: A Pair Of Billion-Dollar Deals For Cyber And AI Infrastructure Lead
AI took five of the ten largest US venture rounds in the week of July 4 to 10, about $3.1 billion combined, led by $1 billion each for machine-identity security firm Keyfactor and chipmaker SambaNova at an $11 billion valuation. Compliance startup Norm AI raised $120 million at a $1.2 billion valuation.
▸ The MSP Angle
Where is AI venture money going in mid-2026?
Machine identity drawing a billion-dollar check is the tell: as agents multiply, the credentials they hold become the attack surface, and the security vendor landscape MSPs buy from will consolidate around that problem. Expect the identity line of your stack to get both better and more expensive over the next 18 months.
Read at Crunchbase News ↗
Thryv · Jul 9