AI Competition Radar: June 28-July 5, 2026

Coverage window: June 28 at 8:23 p.m. to July 5 at 8:00 p.m. Pacific time.

The scouting state improved this week. CISPA published a full European Cybersecurity and AI Hackathon podium, USAII's 2026 Global AI Hackathon winner list is now usable, and UC San Diego's Basement Demo Day produced three ranked student-startup winners. 1 2 3

The other half of the week is still a follow-up queue. Band of Agents, CalHacks AI 2026, Global South AI Safety, Qualcomm x Meta ExecuTorch, and Microsoft Build AI India still lacked complete public winner lists at collection close, so their signals should stay in watchlist or proxy status rather than being treated as final rankings. 4 5 6 7 8

Confirmed results to act on

Event	Confirmed teams	Scout read
CISPA European Cybersecurity and AI Hackathon Championship	SeeWeedZ from Sweden's KTH Royal Institute of Technology won first place and EUR 4,000; Syntax Terror from Poland won second place and EUR 2,000; Advanced Persistent Thinkers from Poland won third place and EUR 1,200. 1	This was a narrow AI-security funnel: 17 finalist teams came from six regional qualifiers and 300 total participants. The technical screen is stronger than a generic app hackathon. 1
USAII Global AI Hackathon 2026	The grand-prize winners were Alyosha in the high-school track, Zen in the undergraduate track, and Opioid Action Engine in the graduate track; USAII also listed runner-up, third-place, Responsible AI, and Social Impact winners in each track. 2	This is the broadest newly usable list: USAII reported 6,000+ pre-registered participants from 90+ countries, 808 qualified teams, 2,284 students, and 622 submitted AI projects. 2
UCSD Basement Demo Day	SunBreak won first place, Patient Compass placed second, and HINDSIGHT placed third at the July 2 Demo Day. 3	This is a student-startup screen rather than a pure AI hackathon. The event had 28 student teams and a prize pool above $65,000, so the best use is targeted founder outreach around the ranked teams and the AI-adjacent portfolio. 3

CISPA: the strongest technical winner this cycle

SeeWeedZ is the cleanest immediate scout lead because the competition format tested AI-security research ability under time pressure. CISPA's finale ran July 2-3 in Saarbrucken, with an awards ceremony at Ingobertusfest in St. Ingbert. 1 The organizer named four SeeWeedZ members: Shekhar Upadhyay, Ankit Grover, Michael Yu, and Edvin Nordqvist. 9

The diligence angle is technical depth. The three finale challenges covered text watermark localization across TextSeal, Gumbel-Max, Unigram, and KGW schemes; member-versus-generated inference against a hidden detector; and federated-learning data reconstruction from gradients across 12 client models and 1,536 private 64 x 64 face images. 1 9 The public gap is commercialization: no GitHub repository, demo link, or company vehicle was found for SeeWeedZ in the supplied research materials. 9

USAII's winner list is broad enough to screen by problem domain. In the high-school track, Alyosha targets reentry support for formerly incarcerated people, Wayfinder targets refugee-benefit navigation, and Cabinet Clear targets discharge-letter and medicine-cabinet interpretation. 2 In the undergraduate track, Zen is a multilingual benefits navigator, TulongAI is a Philippines-focused benefits navigator, and Launchify is an AI co-pilot for early founders. 2 In the graduate track, Opioid Action Engine turns opioid-crisis data into policy action, TowerGuard models the cost of air-traffic-controller shortages, and Team Outliers triages overflow emergency calls. 2

For scouts, the strongest pattern is public-sector navigation and crisis response. Alyosha, Wayfinder, Zen, TulongAI, Opioid Action Engine, TowerGuard, and Team Outliers all sit in workflows where data access, caseworker adoption, and institutional sales will matter more than model novelty. 2 The main blocker is contactability: the research materials list team member names, but they do not provide a complete set of founder social profiles or project repositories. 2

UCSD: ranked startups plus two AI-adjacent teams

SunBreak is the UCSD winner to screen first. The Basement describes SunBreak as a smart energy platform that shifts heavy electrical loads according to on-site generation, grid pricing, and customer energy needs. 10 The listed founders are Spencer Dunn and Sam Andre. 10

The two other ranked teams are less directly tied to AI infrastructure but still relevant. Patient Compass uses tactile 3D models and illustrated guides for doctor-led patient education, with Vanessa Ho listed as founder. 11 HINDSIGHT converts photos, videos, or text descriptions into Unreal Engine-compatible explorable worlds, with Adam Rolander and Anthony Mitine listed as co-founders. 12 The AI-adjacent watchlist beyond the podium is MetaSync, an AI-powered PCOS symptom-tracking app founded by Sharana Sabesan, and Cortexa, an adaptive exoskin project for proactive central-nervous-system protection founded by Karina Bilokur. 13 14

Pending results and proxy signals

Event	Public status	Usable scout signal
Band of Agents	The hackathon ended June 19, and lablab.ai still showed judging in progress with no official winners at collection close. 4	Community votes are only a proxy: Novus had 146 votes, CIPHER AI had 141, AEGIS Fund had 107, and HireMate had 107. 4
CalHacks AI 2026	The event ended June 21, and no complete winner announcement was found; Devpost showed 400 submissions and a winner ribbon on Lucid Voice without a public prize category. 5 15	Treat Lucid Voice as a flagged project, not a ranked winner, until CalHacks publishes award categories. 15
Global South AI Safety	Apart Research's sprint ended June 21, and the public sprint page lists regional tracks across Latin America, Africa, and Asia with $6,000 in prizes plus Apart Fellowship invitations for top teams. 6	Keep this in pending status until regional winners and fellowship invitations are public. 6
Qualcomm x Meta ExecuTorch	The June 27-28 San Francisco event had 100+ participants and 20+ teams, but lablab.ai still described winners as being selected. 7	The category signal is on-device AI: projects used Snapdragon-powered Samsung Galaxy S25 Ultra devices with PyTorch, ExecuTorch, and Snapdragon NPU deployment constraints. 7
Microsoft Build AI India	The HackerEarth event listed a INR 600,000 prize pool, six tracks, and a June final window, but final winners were not publicly accessible in the research materials. 8	The visible finalist list includes Provenant, SellerLens, AgentShield AI, SubstrateOS, Veritrail, ThreatScout, OpenBI, Pramaan, AgentSense, and SwarmOps; treat these as finalists, not final winners. 8

The practical rule is simple: do not compare confirmed winners against pending leaderboards. Band of Agents community votes, CalHacks winner ribbons, and Microsoft Build AI finalist names can help scouts decide what to monitor, but they do not carry the same weight as a published podium. 4 5 8

Corrected result dates

Event	Correct public date	Why scouts should change the calendar
GitLab Transcend	Winners are scheduled for July 10, 2026 at 11:00 a.m. Pacific time. 16	This is now the nearest confirmed Devpost result date in the queue. 16
Mind the Product: Everyone Ships Now	Winners are scheduled for July 15, 2026. 17	The event should stay in next-week monitoring rather than being counted as already judged. 17
H0: Hack the Zero Stack	Winners are scheduled for July 31, 2026. 18	The result is about a month after the June 29 submission deadline, so immediate winner outreach would be premature. 18
UiPath AgentHack	Registration closes July 7, finalists are scheduled for July 15, the live finale is July 23, and winners are scheduled for August 4. 19	The event is still in process, not a missed June result. 19
FIND EVIL!	Winners are scheduled for July 24, 2026 at 9:00 a.m. Pacific time, with judging running June 19-July 17. 20	The prior June 29 assumption should be discarded; the official schedule points to late July. 20
Splunk Agentic Ops	Winners are scheduled for July 17, 2026. 21	This sits in the same mid-July result cluster as GitLab, Mind the Product, and Google Rapid Agent. 21

Next scout calendar: July 6-16

Date	Event	Scout action
July 6-11	AMD Developer Hackathon ACT II starts with 19,409 registered builders, 3,939 teams forming, and a $21,000 prize pool across three tracks plus partner prizes. 22 23	Pre-map teams in token-efficient routing agents, video captioning, and product-focused AI startup concepts before submissions close. 22
July 7	Animoca Brands and HKSTP host the Build East Agentic AI Demo Day in Hong Kong with eight agentic AI startups and a US$10 million investment pool. 24	Follow for Hong Kong agentic AI deal flow; team names were not publicly disclosed in the supplied materials. 24
July 8-9	RAISE Summit runs at Le Carrousel du Louvre in Paris with 10 finalists selected from 1,500+ applicants and tracks sponsored by Backblaze and Google for Startups with DeepMind. 25	Prioritize the final top-three main-stage pitch list once it is public. 25
July 9	ITU AI for Good Innovation Factory Grand Finale runs at Palexpo in Geneva during the AI for Good Global Summit, with Viseur AI and Emily.AI confirmed as regional winners advancing to the finale. 26	Best fit for scouts tracking public-sector, development, and responsible-AI startup pipelines. 26
July 10	Cyber Valley AI Incubator Batch #8 Demo Day runs at Amazon R&D Center in Tubingen, Germany, from 7:00 to 10:45 a.m. Pacific time. 27	Useful for European deep-tech AI scouting; the supplied materials do not include participating team names. 27
July 10-16	MLH Global Hack Week: Season Launch runs digitally worldwide; MLH also lists AI for Social Good with DigitalOcean on July 10-11 in San Francisco. 28	Treat MLH as a student-talent source first and a startup sourcing source second. 28
July 15-16	Google for Startups lists Korea Accelerator Demo Day on July 15 and Brazil Accelerator Demo Day on July 16. 29	Wait for team lists before ranking individual companies; the dates are confirmed, but specific startup rosters were not published in the supplied materials. 29
July 16	Google Cloud Rapid Agent winners are expected at 2:00 p.m. Pacific time, but the supplied research could not independently verify the Devpost details because the page was inaccessible. 30	Keep the event on calendar, but do not quote the participant count or prize amount as independently verified. 30

Outreach order

Contact or map SeeWeedZ first if the mandate includes AI security, watermarking, privacy attacks, or federated-learning risk; the competition challenges are closer to applied research than to generic app-building. 1 9
Build a USAII shortlist by workflow, not by track: benefits navigation, reentry support, opioid-policy intelligence, and emergency triage are the most diligence-ready clusters in the published winner list. 2
Put July 10-17 on active monitoring for GitLab Transcend, Mind the Product, Splunk Agentic Ops, and Google Rapid Agent, while keeping CalHacks and Band of Agents separated as unresolved result gaps. 16 17 21 30

Cover image: CISPA award ceremony image from CISPA.